engram-es_layout_v1.0.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "J1wRG8laa8Pm"
   },
   "source": [
    "# Engram-es Spanish keyboard layout\n",
    "\n",
    "Engram-es is a key layout optimized for comfortable and efficient touch typing in Spanish \n",
    "created by [Arno Klein](https://binarybottle.com) with help from Ian Douglas, aMiguel Guzmán, and Nick Gutiérrez.\n",
    "There is [open source code](https://github.com/binarybottle/engram-es)\n",
    "based on the original English Engram [open source code](https://github.com/binarybottle/engram).\n",
    "<!-- You can install the engram-es layout on [Windows, macOS, and Linux](https://keyman.com/keyboards/engram)\n",
    "or [try it out online](https://keymanweb.com/#en,Keyboard_engram).-->\n",
    "\n",
    "Letters are optimally arranged according to ergonomics factors that promote reduction of lateral finger movements and more efficient typing of high-frequency letter pairs. The most common punctuation marks and special key for diacritical marks (☆) are logically grouped together in the middle columns and numbers are paired with mathematical and logic symbols (shown as pairs of default and Shift-key-accessed characters). See below for a full description.\n",
    "\n",
    "\n",
    "          [ | = ~ +   <  >   ^ & % * ] \\\n",
    "          ( 1 2 3 4   5  6   7 8 9 0 ) /\n",
    "\n",
    "            Z H O X  .:  \"'  M D B J W -_ @#\n",
    "            P I E A  ,;  ☆   T S N R K\n",
    "            F Y Q U  ¿¡  ?!  C L V G\n",
    "            \n",
    "    ☆ + aeiouAEIOU = áéíóúÁÉÍÓÚ (acute accent)\n",
    "    ☆ + nN = ñÑ\n",
    "    ☆ + cC = çÇ\n",
    "    ☆ + Shift         + [letter] = [letter] with a diaresis/umlaut: ü\n",
    "    ☆         + AltGr + [letter] = [letter] with a grave accent: è\n",
    "    ☆ + Shift + AltGr + [letter] = [letter] with a circumflex: â\n",
    "    AltGr + ( = { (open curly brace)\n",
    "    AltGr + ) = } (close curly brace)\n",
    "    AltGr + 5 = « (open quote/comillas) \n",
    "    AltGr + 6 = » (close quote/comillas)\n",
    "    AltGr + - = — (em dash)\n",
    "    AltGr + ' = ` (back tick)\n",
    "    AltGr + . = • (middle dot, or \"interpunct\")\n",
    "    AltGr + s = $ (dollar currency)\n",
    "    AltGr + e = € (euro currency)\n",
    "    AltGr + l = £ (pound currency)\n",
    "\n",
    "### Standard diagonal keyboard (default and Shift-key layers)\n",
    "![Standard keyboard](https://github.com/binarybottle/engram-es/blob/master/assets/engram-es.png?raw=true)\n",
    "\n",
    "(c) 2021 Arno Klein, MIT license\n",
    "\n",
    "----------------"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "awscg4wBa8Po"
   },
   "source": [
    "# Contents\n",
    "1. [Why a new keyboard layout?](#why)\n",
    "2. [Guiding criteria](#criteria)\n",
    "3. Setup:\n",
    "    - [Dependencies and functions](#import)\n",
    "    - [Speed matrix](#speed)\n",
    "    - [Strength matrix](#strength)\n",
    "    - [Flow matrix and Engram scoring model](#flow)\n",
    "4. Steps:\n",
    "    - [Step 1: Define the shape of the key layout to minimize lateral finger movements](#step1)\n",
    "    - [Step 2: Arrange the most frequent letters based on comfort and bigram frequencies](#step2)\n",
    "    - [Step 3: Optimize assignment of the remaining letters](#step3)\n",
    "    - [Step 4: Evaluate winning layout](#step4)\n",
    "    - [Step 5: Arrange non-letter characters in easy-to-remember places](#step5)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "SSdE4O9Wa8Pp"
   },
   "source": [
    "## Why a new keyboard layout? <a name=\"why\">\n",
    "\n",
    "**History** <br>\n",
    "After creating the [English Engram layout](https://engram.dev) ([open source code](https://github.com/binarybottle/engram)), community members came together to help guide the creation of a Spanish version.  Thank you, Nick Gutiérrez (@NickG13) and Miguel Guzmán (@Lobo-Feroz), and a special thanks to Ian Douglas (@iandoug) for cleaning up the Leipzig Spanish corpus and for computing character and bigram frequencies!  For documentation of this Spanish corpus, please see [Creating a Corpus and Chained Bigrams for Spanish Keyboard Development and Evaluation](https://zenodo.org/record/5501931).\n",
    "\n",
    "**Why \"Engram\"?** <br>\n",
    "The name is a pun, referring both to \"n-gram\", letter permutations and their frequencies that are used to compute the Engram layout, and \"engram\", or memory trace, the postulated change in neural tissue to account for the persistence of memory, as a nod to my attempt to make this layout easy to remember."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "wm3T-hmja8Ps"
   },
   "source": [
    "## Guiding criteria   <a name=\"criteria\">\n",
    "\n",
    "    1.  Assign letters to keys that don't require lateral finger movements.\n",
    "    2.  Promote alternating between hands over uncomfortable same-hand transitions.\n",
    "    3.  Assign the most common letters to the most comfortable keys.\n",
    "    4.  Arrange letters so that more frequent bigrams are easier to type.\n",
    "    5.  Promote little-to-index-finger roll-ins over index-to-little-finger roll-outs.\n",
    "    6.  Balance finger loads according to their relative strength.\n",
    "    7.  Avoid stretching shorter fingers up and longer fingers down.\n",
    "    8.  Avoid using the same finger.\n",
    "    9.  Avoid skipping over the home row.\n",
    "    10. Assign the most common punctuation to keys in the middle of the keyboard.\n",
    "    11. Assign easy-to-remember symbols to the Shift-number keys.\n",
    "    \n",
    "### Factors used to compute the Engram layout <a name=\"factors\">\n",
    "  - Spanish letter and letter bigram frequency data from a cleaned-up version of the Leipzig Spanish corpus. The largest file was downloaded from each row of the [original version](https://wortschatz.uni-leipzig.de/en/download/Spanish), except for rows specified as not from Spain, and lines containing non-Spanish names and words were removed.\n",
    "    <br>\n",
    "  - **Flow factors** (transitions between ordered key pairs)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "2eTQ4jxPa8Pv"
   },
   "source": [
    "### Import dependencies and functions  <a name=\"import\">"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# %load code/engram_variables.py\n",
    "# Print .png figures and .txt text files\n",
    "print_output = False # True\n",
    "\n",
    "# Apply strength data\n",
    "apply_strength = True\n",
    "min_strength_factor = 0.9\n",
    "\n",
    "letters24 = ['E','A','O','S','N','I','R','L','D','C','T','U','M',\n",
    "             'P','B','G','V','Q','Y','F','H','J','Z','X','K','W']\n",
    "keys24 = [1,2,3,4, 5,6,7,8, 9,10,11,12, 13,14,15,16, 17,18,19,20, 21,22,23,24]\n",
    "instances24 = [169540585,155261246,114162756,85434855,91757262,83685037,85877478,71380359,63808304,55938235,\n",
    "               56546911,52510166,34957958,34360294,15959118,14964467,12628219,11205355,10437289,10254596,\n",
    "               10147320,5919395,4900008,2224153,982042,584461]\n",
    "max_frequency = 294897235\n",
    "instances_denominator = 100000\n",
    "\n",
    "# Establish which layouts are within a small difference of the top-scoring layout \n",
    "# (the smallest difference between two penalties, 0.9^8 - 0.9^9, in one of 24^2 key pairs):\n",
    "delta = 0.9**8 - 0.9**9\n",
    "factor24 = ((24**2 - 1) + (1-delta)) / (24**2)\n",
    "factor32 = ((32**2 - 1) + (1-delta)) / (32**2)\n",
    "\n",
    "# Establish which layouts are within a small difference of each other when using the speed matrix. \n",
    "# We define an epsilon equal to 13.158 ms for a single bigram (of the 32^2 possible bigrams), \n",
    "# where 13.158 ms is one tenth of 131.58 ms, the fastest measured digraph tapping speed (30,000/228 = 131.58 ms) \n",
    "# recorded in the study: \"Estimation of digraph costs for keyboard layout optimization\", \n",
    "# A Iseri, Ma Eksioglu, International Journal of Industrial Ergonomics, 48, 127-138, 2015.\n",
    "#data_matrix_speed = Speed32x32\n",
    "#time_range = 243  # milliseconds\n",
    "#norm_range = np.max(data_matrix_speed) - np.min(data_matrix_speed)  # 0.6535662299854439\n",
    "#ms_norm = norm_range / time_range  # 0.0026895729629030614\n",
    "#epsilon = 131.58/10 * ms_norm / (32**2)\n",
    "epsilon    = 0.00003549615849447514"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 71
    },
    "colab_type": "code",
    "id": "q1wNgX_FDzRH",
    "outputId": "7c14cebc-a4b7-4a77-d14f-26cbc7690c28"
   },
   "outputs": [],
   "source": [
    "# %load code/engram_functions.py\n",
    "# Import dependencies\n",
    "import xlrd\n",
    "import numpy as np\n",
    "from sympy.utilities.iterables import multiset_permutations\n",
    "import matplotlib\n",
    "import matplotlib.pyplot as plt    \n",
    "import seaborn as sns\n",
    "\n",
    "\n",
    "def permute_optimize_keys(fixed_letters, fixed_letter_indices, open_letter_indices, \n",
    "                          all_letters, keys, data_matrix, bigrams, bigram_frequencies, \n",
    "                          min_score=0, verbose=False):\n",
    "    \"\"\"\n",
    "    Find all permutations of letters, optimize layout, and generate output.\n",
    "    \"\"\"\n",
    "    matrix_selected = select_keys(data_matrix, keys, verbose=False) \n",
    "    \n",
    "    unassigned_letters = []\n",
    "    for all_letter in all_letters:\n",
    "        if all_letter not in fixed_letters:\n",
    "            unassigned_letters.append(all_letter)\n",
    "            if len(unassigned_letters) == len(open_letter_indices):\n",
    "                break\n",
    "\n",
    "    letter_permutations = permute_letters(unassigned_letters, verbose)\n",
    "    if verbose:\n",
    "        print(\"{0} permutations\".format(len(letter_permutations)))\n",
    "    top_permutation, top_score = optimize_layout(np.array([]), matrix_selected, bigrams, bigram_frequencies, \n",
    "                                                 letter_permutations, open_letter_indices, \n",
    "                                                 fixed_letters, fixed_letter_indices, min_score, verbose)\n",
    "    \n",
    "    return top_permutation, top_score, letter_permutations\n",
    "\n",
    "\n",
    "def permute_optimize(starting_permutation, letters, all_letters, all_keys, \n",
    "                     data_matrix, bigrams, bigram_frequencies, min_score=0, verbose=False):\n",
    "    \"\"\"\n",
    "    Find all permutations of letters, optimize layout, and generate output.\n",
    "    \"\"\"\n",
    "    matrix_selected = select_keys(data_matrix, all_keys, verbose=False)\n",
    "    open_positions = []\n",
    "    fixed_positions = [] \n",
    "    open_letters = []\n",
    "    fixed_letters = []\n",
    "    assigned_letters = []\n",
    "    for iletter, letter in enumerate(letters):\n",
    "        if letter.strip() == \"\":\n",
    "            open_positions.append(iletter)\n",
    "            for all_letter in all_letters:\n",
    "                if all_letter not in letters and all_letter not in assigned_letters:\n",
    "                    open_letters.append(all_letter)\n",
    "                    assigned_letters.append(all_letter)\n",
    "                    break\n",
    "        else:\n",
    "            fixed_positions.append(iletter)\n",
    "            fixed_letters.append(letter)\n",
    "\n",
    "    letter_permutations = permute_letters(open_letters, verbose)\n",
    "    if verbose:\n",
    "        print(\"{0} permutations\".format(len(letter_permutations)))\n",
    "    top_permutation, top_score = optimize_layout(starting_permutation, matrix_selected, bigrams, \n",
    "                                                 bigram_frequencies, letter_permutations, open_positions, \n",
    "                                                 fixed_letters, fixed_positions, min_score, verbose)\n",
    "        \n",
    "    return top_permutation, top_score\n",
    "\n",
    "\n",
    "def select_keys(data_matrix, keys, verbose=False):\n",
    "    \"\"\"\n",
    "    Select keys to quantify pairwise relationships.\n",
    "    \"\"\"\n",
    "    # Extract pairwise entries for the keys:\n",
    "    nkeys = len(keys)\n",
    "    Select = np.zeros((nkeys, nkeys))\n",
    "    u = 0\n",
    "    for i in keys:\n",
    "        u += 1\n",
    "        v = 0\n",
    "        for j in keys:\n",
    "            v += 1\n",
    "            Select[u-1,v-1] = data_matrix[i-1,j-1]\n",
    "\n",
    "    # Normalize matrix with min-max scaling to a range with max 1:\n",
    "    newMin = np.min(Select) / np.max(Select)\n",
    "    newMax = 1.0\n",
    "    Select = newMin + (Select - np.min(Select)) * (newMax - newMin) / (np.max(Select) - np.min(Select))\n",
    "    \n",
    "    if verbose:\n",
    "        # Heatmap of array\n",
    "        heatmap(data=Select, title=\"Matrix heatmap\", xlabel=\"Key 1\", ylabel=\"Key 2\", print_output=False); plt.show()\n",
    "    \n",
    "    return Select\n",
    "\n",
    "\n",
    "def permute_letters(letters, verbose=False):\n",
    "    \"\"\"\n",
    "    Find all permutations of a given set of letters (max: 8-10 letters).\n",
    "    \"\"\"\n",
    "    letter_permutations = []\n",
    "    for p in multiset_permutations(letters):\n",
    "        letter_permutations.append(p)\n",
    "    letter_permutations = np.array(letter_permutations)\n",
    "    \n",
    "    return letter_permutations\n",
    "\n",
    "\n",
    "def score_layout(data_matrix, letters, bigrams, bigram_frequencies, verbose=False):\n",
    "    \"\"\"\n",
    "    Compute the score for a given letter-key layout (NOTE normalization step).\n",
    "    \"\"\"\n",
    "    # Create a matrix of bigram frequencies:\n",
    "    nletters = len(letters)\n",
    "    F2 = np.zeros((nletters, nletters))\n",
    "\n",
    "    # Find the bigram frequency for each ordered pair of letters in the permutation:\n",
    "    for i1 in range(nletters):\n",
    "        for i2 in range(nletters):\n",
    "            bigram = letters[i1] + letters[i2]\n",
    "            i2gram = np.where(bigrams == bigram)\n",
    "            if np.size(i2gram) > 0:\n",
    "                F2[i1, i2] = bigram_frequencies[i2gram][0]\n",
    "\n",
    "    # Normalize matrices with min-max scaling to a range with max 1:\n",
    "    newMax = 1\n",
    "    minF2 = np.min(F2)\n",
    "    maxF2 = np.max(F2)\n",
    "    newMin2 = minF2 / maxF2\n",
    "    F2 = newMin + (F2 - minF2) * (newMax - newMin2) / (maxF2 - minF2)\n",
    "\n",
    "    # Compute the score for this permutation:\n",
    "    score  = np.average(data_matrix * F2) \n",
    "    if verbose:\n",
    "        print(\"Score for letter permutation {0}: {1}\".format(letters, score))\n",
    "\n",
    "    return score\n",
    "\n",
    "\n",
    "def tally_bigrams(input_text, bigrams, normalize=True, verbose=False):\n",
    "    \"\"\"\n",
    "    Compute the score for a given letter-key layout (NOTE normalization step).\n",
    "    \"\"\"   \n",
    "    # Find the bigram frequency for each ordered pair of letters in the input text\n",
    "    #input_text = [str.upper(str(x)) for x in input_text]\n",
    "    input_text = [str.upper(x) for x in input_text]\n",
    "    nchars = len(input_text)\n",
    "    F = np.zeros(len(bigrams))\n",
    "\n",
    "    for ichar in range(0, nchars-1):\n",
    "        bigram = input_text[ichar] + input_text[ichar + 1]\n",
    "        i2gram = np.where(bigrams == bigram)\n",
    "        if np.size(i2gram) > 0:\n",
    "            F[i2gram] += 1\n",
    "\n",
    "    # Normalize matrix with min-max scaling to a range with max 1:\n",
    "    if normalize:\n",
    "        newMax = 1\n",
    "        newMin = np.min(F) / np.max(F)\n",
    "        F = newMin + (F - np.min(F)) * (newMax - newMin) / (np.max(F) - np.min(F))\n",
    "\n",
    "    bigram_frequencies_for_input = F\n",
    "\n",
    "    if verbose:\n",
    "        print(\"Bigram frequencies for input: {0}\".format(bigram_frequencies_for_input))\n",
    "\n",
    "    return bigram_frequencies_for_input\n",
    "\n",
    "\n",
    "def tally_layout_samefinger_bigrams(layout, bigrams, bigram_frequencies, nkeys=32, verbose=False):\n",
    "    \"\"\"\n",
    "    Tally the number of same-finger bigrams within (a list of 24 letters representing) a layout:\n",
    "    ['P','Y','O','U','C','I','E','A','G','K','J','X','M','D','L','B','R','T','N','S','H','V','W','F']\n",
    "    \"\"\"  \n",
    "    if nkeys == 32:\n",
    "        #        Left:            Right:\n",
    "        #    1  2  3  4 25   28 13 14 15 16 31 \n",
    "        #    5  6  7  8 26   29 17 18 19 20 32\n",
    "        #    9 10 11 12 27   30 21 22 23 24\n",
    "        same_finger_keys = [[1,5],[5,9],[1,9], [2,6],[6,10],[2,10], \n",
    "                            [3,7],[7,11],[3,11], [4,8],[8,12],[4,12],\n",
    "                            [25,26],[26,27],[25,27], [28,29],[29,30],[28,30], [31,32],\n",
    "                            [4,25],[4,26],[4,27], [8,25],[8,26],[8,27], [12,25],[12,26],[12,27],\n",
    "                            [13,28],[13,29],[13,30], [17,28],[17,29],[17,30], [21,28],[21,29],[21,30],\n",
    "                            [31,16],[31,20],[31,24], [32,16],[32,20],[32,24],\n",
    "                            [13,17],[17,21],[13,21], [14,18],[18,22],[14,22], \n",
    "                            [15,19],[19,23],[15,23], [16,20],[20,24],[16,24]] \n",
    "    elif nkeys == 24:\n",
    "        #    1  2  3  4         13 14 15 16  \n",
    "        #    5  6  7  8         17 18 19 20 \n",
    "        #    9 10 11 12         21 22 23 24\n",
    "        same_finger_keys = [[1,5],[5,9],[1,9], [2,6],[6,10],[2,10], \n",
    "                            [3,7],[7,11],[3,11], [4,8],[8,12],[4,12],\n",
    "                            [13,17],[17,21],[13,21], [14,18],[18,22],[14,22], \n",
    "                            [15,19],[19,23],[15,23], [16,20],[20,24],[16,24]] \n",
    "\n",
    "    layout = [str.upper(x) for x in layout]\n",
    "    max_frequency = 1.00273E+11\n",
    "\n",
    "    samefinger_bigrams = []\n",
    "    samefinger_bigram_counts = []\n",
    "    for bigram_keys in same_finger_keys:\n",
    "        bigram1 = layout[bigram_keys[0]-1] + layout[bigram_keys[1]-1]\n",
    "        bigram2 = layout[bigram_keys[1]-1] + layout[bigram_keys[0]-1]\n",
    "        i2gram1 = np.where(bigrams == bigram1)\n",
    "        i2gram2 = np.where(bigrams == bigram2)\n",
    "        if np.size(i2gram1) > 0:\n",
    "            samefinger_bigrams.append(bigram1)\n",
    "            samefinger_bigram_counts.append(max_frequency * bigram_frequencies[i2gram1] / np.max(bigram_frequencies))\n",
    "        if np.size(i2gram2) > 0:\n",
    "            samefinger_bigrams.append(bigram2)\n",
    "            samefinger_bigram_counts.append(max_frequency * bigram_frequencies[i2gram2] / np.max(bigram_frequencies))\n",
    "\n",
    "    samefinger_bigrams_total = np.sum([x[0] for x in samefinger_bigram_counts])\n",
    "\n",
    "    if verbose:\n",
    "        print(\"    Total same-finger bigram frequencies: {0:15.0f}\".format(samefinger_bigrams_total))\n",
    "\n",
    "    return samefinger_bigrams, samefinger_bigram_counts, samefinger_bigrams_total \n",
    "\n",
    "\n",
    "def tally_layout_bigram_rolls(layout, bigrams, bigram_frequencies, nkeys=32, verbose=False):\n",
    "    \"\"\"\n",
    "    Tally the number of bigrams that engage little-to-index finger inward rolls\n",
    "    for (a list of 24 or 32 letters representing) a layout,\n",
    "    within the four columns of one hand, or any column across two hands.\n",
    "    layout = ['P','Y','O','U','C','I','E','A','G','K','J','X','L','D','B','V','N','T','R','S','H','M','W','F']\n",
    "    bigram_rolls, bigram_roll_counts, bigram_rolls_total = tally_layout_bigram_rolls(layout, bigrams, bigram_frequencies, nkeys=24, verbose=True)\n",
    "    \"\"\"   \n",
    "    if nkeys == 32:\n",
    "        #        Left:            Right:\n",
    "        #    1  2  3  4 25   28 13 14 15 16 31 \n",
    "        #    5  6  7  8 26   29 17 18 19 20 32\n",
    "        #    9 10 11 12 27   30 21 22 23 24\n",
    "\n",
    "        roll_keys = [[1,2],[2,3],[3,4], [5,6],[6,7],[7,8], [9,10],[10,11],[11,12],\n",
    "                    [16,15],[15,14],[14,13], [20,19],[19,18],[18,17], [24,23],[23,22],[22,21],\n",
    "                    [1,3],[2,4],[1,4], [5,7],[6,8],[5,8], [9,11],[10,12],[9,12],\n",
    "                    [16,14],[15,13],[16,13], [20,18],[19,17],[20,17], [24,22],[23,21],[24,21],\n",
    "                    [1,6],[1,7],[1,8],[2,7],[2,8],[3,8], \n",
    "                    [5,2],[5,3],[5,4],[6,3],[6,4],[7,4],\n",
    "                    [5,10],[5,11],[5,12],[6,11],[6,12],[7,12], \n",
    "                    [9,6],[9,7],[9,8],[10,7],[10,8],[11,8],\n",
    "                    [16,19],[16,18],[16,17],[15,18],[15,17],[14,17], \n",
    "                    [20,15],[20,14],[20,13],[19,14],[19,13],[18,13],\n",
    "                    [20,23],[20,22],[20,21],[19,22],[19,21],[18,21], \n",
    "                    [24,19],[24,18],[24,17],[23,18],[23,17],[22,17],\n",
    "                    [1,10],[1,11],[1,12],[2,11],[2,12],[3,12],\n",
    "                    [9,2],[9,3],[9,4],[10,3],[10,4],[11,4],\n",
    "                    [16,23],[16,22],[16,21],[15,22],[15,21],[14,21],\n",
    "                    [24,15],[24,14],[24,13],[23,14],[23,13],[22,13]]\n",
    "        for i in [1,2,3,4,5,6,7,8,9,10,11,12, 25,26,27]:\n",
    "            for j in [13,14,15,16,17,18,19,20,21,22,23,24, 28,29,30,31,32]:\n",
    "                roll_keys.append([i,j])\n",
    "        for i in [13,14,15,16,17,18,19,20,21,22,23,24, 28,29,30,31,32]:\n",
    "            for j in [1,2,3,4,5,6,7,8,9,10,11,12, 25,26,27]:\n",
    "                roll_keys.append([i,j])\n",
    "\n",
    "    elif nkeys == 24:\n",
    "        #    1  2  3  4         13 14 15 16  \n",
    "        #    5  6  7  8         17 18 19 20 \n",
    "        #    9 10 11 12         21 22 23 24\n",
    "        roll_keys = [[1,2],[2,3],[3,4], [5,6],[6,7],[7,8], [9,10],[10,11],[11,12],\n",
    "                    [16,15],[15,14],[14,13], [20,19],[19,18],[18,17], [24,23],[23,22],[22,21],\n",
    "                    [1,3],[2,4],[1,4], [5,7],[6,8],[5,8], [9,11],[10,12],[9,12],\n",
    "                    [16,14],[15,13],[16,13], [20,18],[19,17],[20,17], [24,22],[23,21],[24,21],\n",
    "                    [1,6],[1,7],[1,8],[2,7],[2,8],[3,8], [5,2],[5,3],[5,4],[6,3],[6,4],[7,4],\n",
    "                    [5,10],[5,11],[5,12],[6,11],[6,12],[7,12], [9,6],[9,7],[9,8],[10,7],[10,8],[11,8],\n",
    "                    [16,19],[16,18],[16,17],[15,18],[15,17],[14,17], [20,15],[20,14],[20,13],[19,14],[19,13],[18,13],\n",
    "                    [20,23],[20,22],[20,21],[19,22],[19,21],[18,21], [24,19],[24,18],[24,17],[23,18],[23,17],[22,17],\n",
    "                    [1,10],[1,11],[1,12],[2,11],[2,12],[3,12], [9,2],[9,3],[9,4],[10,3],[10,4],[11,4],\n",
    "                    [16,23],[16,22],[16,21],[15,22],[15,21],[14,21], [24,15],[24,14],[24,13],[23,14],[23,13],[22,13]]\n",
    "        for i in range(0,12):\n",
    "            for j in range(12,24):\n",
    "                roll_keys.append([i,j])\n",
    "        for i in range(12,24):\n",
    "            for j in range(0,12):\n",
    "                roll_keys.append([i,j])\n",
    "\n",
    "    layout = [str.upper(x) for x in layout]\n",
    "    max_frequency = 1.00273E+11\n",
    "\n",
    "    bigram_rolls = []\n",
    "    bigram_roll_counts = []\n",
    "    for bigram_keys in roll_keys:\n",
    "        bigram1 = layout[bigram_keys[0]-1] + layout[bigram_keys[1]-1]\n",
    "        bigram2 = layout[bigram_keys[1]-1] + layout[bigram_keys[0]-1]\n",
    "        i2gram1 = np.where(bigrams == bigram1)\n",
    "        i2gram2 = np.where(bigrams == bigram2)\n",
    "        if np.size(i2gram1) > 0:\n",
    "            bigram_rolls.append(bigram1)\n",
    "            bigram_roll_counts.append(max_frequency * bigram_frequencies[i2gram1] / np.max(bigram_frequencies))\n",
    "        if np.size(i2gram2) > 0:\n",
    "            bigram_rolls.append(bigram2)\n",
    "            bigram_roll_counts.append(max_frequency * bigram_frequencies[i2gram2] / np.max(bigram_frequencies))\n",
    "\n",
    "    bigram_rolls_total = np.sum([x[0] for x in bigram_roll_counts])\n",
    "\n",
    "    if verbose:\n",
    "        print(\"    Total bigram inward roll frequencies: {0:15.0f}\".format(bigram_rolls_total))\n",
    "\n",
    "    return bigram_rolls, bigram_roll_counts, bigram_rolls_total \n",
    "\n",
    "\n",
    "def optimize_layout(starting_permutation, data_matrix, bigrams, bigram_frequencies, letter_permutations, \n",
    "                    open_positions, fixed_letters, fixed_positions=[], min_score=0, verbose=False):\n",
    "    \"\"\"\n",
    "    Compute scores for all letter-key layouts.\n",
    "    \"\"\"\n",
    "    top_permutation = starting_permutation\n",
    "    top_score = min_score\n",
    "    use_score_function = False\n",
    "\n",
    "    nletters = len(open_positions) + len(fixed_positions)\n",
    "    F2 = np.zeros((nletters, nletters))\n",
    "\n",
    "    # Loop through the permutations of the selected letters:\n",
    "    for p in letter_permutations:\n",
    "        letters = np.array(['W' for x in range(nletters)])  # KEEP to initialize!\n",
    "        for imove, open_position in enumerate(open_positions):\n",
    "            letters[open_position] = p[imove]\n",
    "        for ifixed, fixed_position in enumerate(fixed_positions):\n",
    "            letters[fixed_position] = fixed_letters[ifixed]\n",
    "\n",
    "        # Compute the score for this permutation:\n",
    "        if use_score_function:\n",
    "            score = score_layout(data_matrix, letters, bigrams, bigram_frequencies, verbose=False)\n",
    "        else:\n",
    "            # Find the bigram frequency for each ordered pair of letters in the permutation:\n",
    "            for i1 in range(nletters):\n",
    "                for i2 in range(nletters):\n",
    "                    bigram = letters[i1] + letters[i2]\n",
    "                    i2gram = np.where(bigrams == bigram)\n",
    "                    if np.size(i2gram) > 0:\n",
    "                        F2[i1, i2] = bigram_frequencies[i2gram][0]\n",
    "\n",
    "            # Normalize matrices with min-max scaling to a range with max 1:\n",
    "            newMax = 1\n",
    "            minF2 = np.min(F2)\n",
    "            maxF2 = np.max(F2)\n",
    "            newMin2 = minF2 / maxF2\n",
    "            F = newMin + (F2 - minF2) * (newMax - newMin2) / (maxF2 - minF2)\n",
    "\n",
    "            # Compute the score for this permutation:\n",
    "            score  = np.average(data_matrix * F) \n",
    "\n",
    "        if score > top_score:\n",
    "            top_score = score\n",
    "            top_permutation = letters\n",
    "            \n",
    "    if verbose:\n",
    "        if top_score == min_score:\n",
    "            print(\"top_score = min_score\")\n",
    "        print(\"{0:0.8f}\".format(top_score))\n",
    "        print(*top_permutation)\n",
    "        \n",
    "    return top_permutation, top_score\n",
    "\n",
    "\n",
    "def exchange_letters(letters, fixed_letter_indices, all_letters, all_keys, data_matrix, \n",
    "                     bigrams, bigram_frequencies, verbose=True):\n",
    "    \"\"\"\n",
    "    Exchange letters, 8 keys at a time (8! = 40,320) selected twice in 14 different ways:\n",
    "\n",
    "    Indices:\n",
    "         0  1  2  3     12 13 14 15\n",
    "         4  5  6  7     16 17 18 19\n",
    "         8  9 10 11     20 21 22 23 \n",
    "\n",
    "    1. Top rows\n",
    "         0  1  2  3     12 13 14 15\n",
    "    2. Bottom rows\n",
    "         8  9 10 11     20 21 22 23 \n",
    "    3. Top and bottom rows on the right side\n",
    "                        12 13 14 15\n",
    "                        20 21 22 23 \n",
    "    4. Top and bottom rows on the left side \n",
    "         0  1  2  3\n",
    "         8  9 10 11 \n",
    "    5. Top right and bottom left rows\n",
    "                        12 13 14 15\n",
    "         8  9 10 11 \n",
    "    6. Top left and bottom right rows \n",
    "         0  1  2  3\n",
    "                        20 21 22 23 \n",
    "    7. Center of the top and bottom rows on both sides\n",
    "            1  2          13 14\n",
    "            9 10          21 22\n",
    "    8. The eight corners\n",
    "         0        3    12       15\n",
    "         8       11    20       23 \n",
    "    9. Left half of the top and bottom rows on both sides \n",
    "         0  1          12 13\n",
    "         8  9          20 21\n",
    "    10. Right half of the top and bottom rows on both sides\n",
    "               2  3          14 15\n",
    "              10 11          22 23 \n",
    "    11. Left half of non-home rows on the left and right half of the same rows on the right \n",
    "         0  1                14 15\n",
    "         8  9                22 23 \n",
    "    12. Right half of non-home rows on the left and left half of the same rows on the right\n",
    "               2  3    12 13\n",
    "              10 11    20 21 \n",
    "    13. Top center and lower sides\n",
    "            1  2           13 14\n",
    "         8       11     20       23 \n",
    "    14. Top sides and lower center\n",
    "         0        3     12       15\n",
    "            9 10           21 22   \n",
    "    15. Repeat 1-14\n",
    "\n",
    "    \"\"\"\n",
    "    top_score = score_layout(data_matrix, letters, bigrams, bigram_frequencies, verbose=False)\n",
    "    print('Initial score: {0}'.format(top_score)) \n",
    "    print(*letters) \n",
    "    top_permutation = letters\n",
    "\n",
    "    lists_of_open_indices = [\n",
    "        [0,1,2,3,12,13,14,15],\n",
    "        [8,9,10,11,20,21,22,23],\n",
    "        [12,13,14,15,20,21,22,23],\n",
    "        [0,1,2,3,8,9,10,11],\n",
    "        [12,13,14,15,8,9,10,11],\n",
    "        [0,1,2,3,20,21,22,23],\n",
    "        [1,2,13,14,9,10,21,22],\n",
    "        [0,3,12,15,8,11,20,23],\n",
    "        [0,1,12,13,8,9,20,21],\n",
    "        [2,3,14,15,10,11,22,23],\n",
    "        [0,1,14,15,8,9,22,23],\n",
    "        [2,3,12,13,10,11,20,21],\n",
    "        [1,2,8,11,13,14,20,23],\n",
    "        [0,3,9,10,12,15,21,22]  \n",
    "    ]\n",
    "    lists_of_print_statements = [\n",
    "        '1. Top rows',\n",
    "        '2. Bottom rows',\n",
    "        '3. Top and bottom rows on the right side',\n",
    "        '4. Top and bottom rows on the left side',\n",
    "        '5. Top right and bottom left rows',\n",
    "        '6. Top left and bottom right rows',\n",
    "        '7. Center of the top and bottom rows on both sides',\n",
    "        '8. The eight corners',\n",
    "        '9. Left half of the top and bottom rows on both sides',\n",
    "        '10. Right half of the top and bottom rows on both sides',\n",
    "        '11. Left half of non-home rows on the left and right half of the same rows on the right',\n",
    "        '12. Right half of non-home rows on the left and left half of the same rows on the right',\n",
    "        '13. Top center and lower sides',\n",
    "        '14. Top sides and lower center'\n",
    "    ]\n",
    "                         \n",
    "    for istep in [1,2]:\n",
    "        if istep == 1:\n",
    "            s = \"Set 1: 14 letter exchanges: \"\n",
    "        elif istep == 2:\n",
    "            s = \"Set 2: 14 letter exchanges: \"\n",
    "\n",
    "        for ilist, open_indices in enumerate(lists_of_open_indices):\n",
    "            print_statement = lists_of_print_statements[ilist]     \n",
    "\n",
    "            if verbose:\n",
    "                print('{0} {1}'.format(s, print_statement))\n",
    "              \n",
    "            starting_permutation = top_permutation.copy()\n",
    "            for open_index in open_indices:\n",
    "                if open_index not in fixed_letter_indices:\n",
    "                    top_permutation[open_index] = ''\n",
    "            \n",
    "            top_permutation, top_score = permute_optimize(starting_permutation, top_permutation, letters24, \n",
    "                                                          keys24, data_matrix, bigrams, bigram_frequencies, \n",
    "                                                          min_score=top_score, verbose=True)                    \n",
    "    if verbose:\n",
    "        print('')\n",
    "        print('    -------- DONE --------') \n",
    "        print('')\n",
    "\n",
    "    return top_permutation, top_score\n",
    "                             \n",
    "\n",
    "def rank_within_epsilon(numbers, epsilon, factor=False, verbose=True):\n",
    "    \"\"\"\n",
    "    numbers = np.array([10,9,8,7,6])\n",
    "    epsilon = 1\n",
    "    rank_within_epsilon(numbers, epsilon, factor=False, verbose=True) \n",
    "    >>> array([1., 1., 2., 2., 3.])\n",
    "    numbers = np.array([0.798900824, 0.79899900824, 0.79900824])\n",
    "    epsilon = 0.9**8 - 0.9**9\n",
    "    factor24 = ((24**2 - 1) + (1-epsilon)) / (24**2) # 0.999925266109375\n",
    "    rank_within_epsilon(numbers, factor24, factor=True, verbose=True) \n",
    "    >>> array([2., 1., 1.])\n",
    "    \"\"\"\n",
    "    numbers = np.array(numbers)\n",
    "    Isort = np.argsort(-numbers)\n",
    "    numbers_sorted = numbers[Isort]\n",
    "    count = 1\n",
    "    ranks = np.zeros(np.size(numbers))\n",
    "    for i, num in enumerate(numbers_sorted):\n",
    "        if ranks[i] == 0:\n",
    "            if factor:\n",
    "                lower_bound = num * epsilon\n",
    "            else:\n",
    "                lower_bound = num - epsilon\n",
    "            bounded_nums1 = num >= numbers_sorted\n",
    "            bounded_nums2 = numbers_sorted >= lower_bound\n",
    "            bounded_nums = bounded_nums1 * bounded_nums2\n",
    "            count += 1\n",
    "            for ibounded, bounded_num in enumerate(bounded_nums):\n",
    "                if bounded_num == True:\n",
    "                    ranks[ibounded] = count\n",
    "\n",
    "    uranks = np.unique(ranks)\n",
    "    nranks = np.size(uranks)\n",
    "    new_ranks = ranks.copy()\n",
    "    new_count = 0\n",
    "    for rank in uranks:\n",
    "        new_count += 1\n",
    "        same_ranks = ranks == rank\n",
    "        for isame, same_rank in enumerate(same_ranks):\n",
    "            if same_rank == True:\n",
    "                new_ranks[isame] = new_count\n",
    "\n",
    "    #ranks_sorted = new_ranks[Isort]\n",
    "    ranks_sorted = [np.int(x) for x in new_ranks]\n",
    "    \n",
    "    if verbose:\n",
    "        for i, num in enumerate(numbers_sorted):\n",
    "            print(\"    ({0})    {1}\".format(np.int(ranks_sorted[i]), num))\n",
    "        \n",
    "    return numbers_sorted, ranks_sorted, Isort\n",
    "\n",
    "\n",
    "def print_matrix_info(matrix_data, matrix_label, nkeys, nlines=10):\n",
    "    \"\"\"\n",
    "    Print matrix output.\n",
    "    \"\"\"\n",
    "    print(\"{0} min = {1}, max = {2}\".format(matrix_label, np.min(matrix_data), np.max(matrix_data)))\n",
    "    matrix_flat = matrix_data.flatten()\n",
    "    argsort = np.argsort(matrix_flat)\n",
    "    print(\"{0} key number pairs with minimum values:\".format(matrix_label))\n",
    "    for x in argsort[0:nlines]:\n",
    "        if x % nkeys == 0:\n",
    "            min_row = np.int(np.ceil(x / nkeys)) + 1\n",
    "            min_col = 1\n",
    "        else:\n",
    "            min_row = np.int(np.ceil(x / nkeys))\n",
    "            min_col = x - nkeys * (min_row-1) + 1                \n",
    "        print(\"        {0} -> {1}        ({2})\".format(min_row, min_col, matrix_flat[x]))\n",
    "    print(\"{0} key number pairs with maximum values:\".format(matrix_label))\n",
    "    max_sort = argsort[-nlines::]\n",
    "    for x in max_sort[::-1]:\n",
    "        if x % nkeys == 0:\n",
    "            max_row = np.int(np.ceil(x / nkeys)) + 1\n",
    "            max_col = 1\n",
    "        else:\n",
    "            max_row = np.int(np.ceil(x / nkeys))\n",
    "            max_col = x - nkeys * (max_row-1) + 1                \n",
    "        print(\"        {0} -> {1}        ({2})\".format(max_row, max_col, matrix_flat[x]))\n",
    "\n",
    "\n",
    "def heatmap(data, title=\"\", xlabel=\"\", ylabel=\"\", x_axis_labels=[], y_axis_labels=[], print_output=True):\n",
    "    \"\"\"\n",
    "    Plot heatmap of matrix.\n",
    "    \"\"\"\n",
    "    # use heatmap function, set the color as viridis and\n",
    "    # make each cell seperate using linewidth parameter\n",
    "    plt.figure()\n",
    "    sns_plot = sns.heatmap(data, xticklabels=x_axis_labels, yticklabels=y_axis_labels, linewidths=1, \n",
    "                           cmap=\"viridis\", square=True, vmin=np.min(data), vmax=np.max(data))\n",
    "    plt.title(title)\n",
    "    plt.xlabel(xlabel)\n",
    "    plt.ylabel(ylabel)\n",
    "    sns_plot.set_xticklabels(x_axis_labels)  #, rotation=75) \n",
    "    sns_plot.set_yticklabels(y_axis_labels) \n",
    "    if print_output:\n",
    "        sns_plot.figure.savefig(\"{0}_heatmap.png\".format(title))\n",
    "        \n",
    "    \n",
    "def histmap(data, title=\"\", print_output=True):\n",
    "    \"\"\"\n",
    "    Plot histogram.\n",
    "    \"\"\"\n",
    "    sns.distplot(data)\n",
    "    plt.title(title)\n",
    "    if print_output:\n",
    "        sns_plot.figure.savefig(\"{0}_histogram.png\".format(title))\n",
    "        \n",
    "        \n",
    "def print_layout24(layout):\n",
    "    \"\"\"\n",
    "    Print layout.\n",
    "    \"\"\"   \n",
    "    print('    {0}  {1}'.format(' '.join(layout[0:4]),\n",
    "                                ' '.join(layout[12:16])))\n",
    "    print('    {0}  {1}'.format(' '.join(layout[4:8]),\n",
    "                                ' '.join(layout[16:20])))\n",
    "    print('    {0}  {1}'.format(' '.join(layout[8:12]),\n",
    "                                ' '.join(layout[20:24])))\n",
    "\n",
    "\n",
    "def print_layout24_instances(layout, letters24, instances24, bigrams, bigram_frequencies):\n",
    "    \"\"\"\n",
    "    Print billions of instances per letter (not Z or Q) in layout form.\n",
    "    layout = ['P','Y','O','U','C','I','E','A','G','K','J','X','M','D','L','B','R','T','N','S','H','V','W','F']\n",
    "    print_layout24_instances(layout, letters24, instances24, bigrams, bigram_frequencies)\n",
    "    \"\"\"\n",
    "    layout_instances = []\n",
    "    layout_instances_strings = []\n",
    "    for letter in layout:\n",
    "        index = letters24.index(letter)\n",
    "        layout_instances.append(instances24[index])\n",
    "        layout_instances_strings.append('{0:3.0f}'.format(instances24[index]/instances_denominator))\n",
    " \n",
    "    print('    {0}  {1}'.format(' '.join(layout_instances_strings[0:4]),\n",
    "                                ' '.join(layout_instances_strings[12:16])))\n",
    "    print('    {0}  {1}'.format(' '.join(layout_instances_strings[4:8]),\n",
    "                                ' '.join(layout_instances_strings[16:20])))\n",
    "    print('    {0}  {1}'.format(' '.join(layout_instances_strings[8:12]),\n",
    "                                ' '.join(layout_instances_strings[20:24])))\n",
    "    left_sum = np.sum(layout_instances[0:12])\n",
    "    right_sum = np.sum(layout_instances[12:24])\n",
    "    pL = ''\n",
    "    pR = ''\n",
    "    if left_sum > right_sum:\n",
    "        pL = ' ({0:3.2f}%)'.format(100 * (left_sum - right_sum) / right_sum)\n",
    "    elif right_sum > left_sum:\n",
    "        pR = ' ({0:3.2f}%)'.format(100 * (right_sum - left_sum) / left_sum)\n",
    "    \n",
    "    print('\\n    left: {0}{1}  right: {2}{3}'.format(left_sum, pL, right_sum, pR))\n",
    "    \n",
    "    tally_layout_samefinger_bigrams(layout, bigrams, bigram_frequencies, nkeys=24, verbose=True)\n",
    "    tally_layout_bigram_rolls(layout, bigrams, bigram_frequencies, nkeys=24, verbose=True)\n",
    "   \n",
    "\n",
    "def print_bigram_frequency(input_pair, bigrams, bigram_frequencies):\n",
    "    \"\"\"\n",
    "    >>> print_bigram_frequency(['t','h'], bigrams, bigram_frequencies)\n",
    "    \"\"\"\n",
    "    # Find the bigram frequency\n",
    "    input_text = [str.upper(str(x)) for x in input_pair]\n",
    "    nchars = len(input_text)\n",
    "    for ichar in range(0, nchars-1):\n",
    "        bigram1 = input_text[ichar] + input_text[ichar + 1]\n",
    "        bigram2 = input_text[ichar + 1] + input_text[ichar]\n",
    "        i2gram1 = np.where(bigrams == bigram1)\n",
    "        i2gram2 = np.where(bigrams == bigram2)\n",
    "        if np.size(i2gram1) > 0:\n",
    "            freq1 = bigram_frequencies[i2gram1[0][0]]\n",
    "            print(\"{0}: {1:3.2f}\".format(bigram1, freq1))\n",
    "        if np.size(i2gram2) > 0:\n",
    "            freq2 = bigram_frequencies[i2gram2[0][0]]\n",
    "            print(\"{0}: {1:3.2f}\".format(bigram2, freq2))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "rFiySi8rDzRN"
   },
   "source": [
    "### Bigram frequencies <a name=\"ngrams\">\n",
    "\n",
    "[NOTE: If you want to compute an optimized layout for another language, or based on another corpus, you can run the tally_bigrams() function above and replace bigram_frequencies below before running the rest of the code.]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "bigrams = np.array(['DE','EN','ES','LA','ER','OS','ON','RA','UE','AS','AN','EL','AR','RE','CO','NT','DO','TA','QU','TE','CI','OR','AL','AD','IO','ST','SE','RO','LO','TO','NA','IA','NO','UN','CA','RI','IN','MA','DA','IE','LE','PA','PO','TI','TR','SI','ME','DI','ND','IC','AC','MO','ID','EC','SA','AB','IS','SO','MI','SU','NE','PR','PE','HA','LI','OM','AM','CE','NC','BA','NI','LL','CU','VI','EM','VE','OL','IT','RT','ED','AT','IR','IM','BR','BI','GU','MP','IL','US','CH','EG','GA','OC','UR','UA','TU','UI','PU','VA','GO','OD','HO','SP','MU','NS','ET','RR','ZA','FI','IG','MB','OB','GR','BL','OT','JO','UL','CT','RM','SC','PI','BE','IV','VO','UC','FU','EA','AP','LU','NU','EV','HE','CR','RD','RS','AG','FE','PL','GE','EX','JA','UD','IB','HI','EJ','EZ','RC','JE','NG','LT','DU','GI','RU','AY','AU','UM','BO','AJ','DR','FA','FO','IZ','EP','OP','JU','RN','EB','ZO','UT','RG','YO','AV','UB','BU','FR','CL','AZ','IP','YA','SM','NF','IF','OV','EO','OG','AI','UY','RL','LG','EF','IJ','AQ','UP','NV','CC','OF','OY','HU','EI','NZ','XI','LM','AF','UG','RV','LV','RQ','YE','XP','LD','AH','UJ','SD','OJ','GL','NQ','UV','EU','TH','XT','EQ','EY','FL','IU','GN','OZ','LC','RP','AE','EE','RB','PT','VU','UZ','BS','RZ','IQ','SL','SF','NM','OI','OE','UO','OU','XC','LS','NR','NJ','YU','II','SS','UF','ZU','SH','JI','OX','RF','DM','LP','OO','OA','PC','SG','KI','LB','SQ','OH','LQ','WA','ZC','BJ','XA','NN','TT','OQ','KE','MN','AO','WI','CK','CN','EH','LF','PS','WE','KA','RY','AX','OW','LY','DV','LZ','TS','BT','SR','TL','RK','PP','NL','ZQ','SN','AK','IX','ZE','SB','UQ','TC','ZG','TB','RJ','SV','NY','KO','OK','GH','NH','TM','LR','WO','ZI','XO','MS','DQ','NK','IK','TY','PH','FF','XX','HN','XV','EW','GM','MM','WH','SK','DS','BY','YS','YN','FT','UX','XE','DY','TW','HR','ZM','AA','TZ','HT','XU','GT','DD','AW','UU','BD','XH','VD','MY','KS','BC','YL','KU','EK','BB','YR','GD','DJ','GS','LK','LN','SY','TN','DH','LH','IH','DL','ZZ','WN','BV','CS','CD','KY','NB','WS','KL','KM','YM','TV','GG','UK','WR','HS','DN','UH','ZN','MC','HY','ML','LX','CY','DG','DW','HM','CM','RH','HL','KN','WW','KR','SW','YD','MR','PY','DT','NP','FM','MT','YI','PN','VR','KH','YT','GY','FC','XQ','PM','BM','RW','PD','VS','BN','YC','ZT','YP','GB','IY','TP','CQ','CF','FP','FS','DC','FG','PF','DF','CP','LJ','YW','TX','CV','VN','ZY','ZB','QA','XL','TF','XM','CB','GP','CG','ZH','PV','PG','ZP','RX','MD','PB','YB','WL','DP','VL','WT','MW','ZL','LW','HD','TG','KK','FN','PJ','NW','HW','IW','DB','BP','XY','KG','SJ','GK','ZK','KB','HC','VY','TD','JR','MV','BH','KW','YV','HB','NX','FY','FB','MG','HF','MX','KT','VT','XF','HP','CZ','WK','SZ','YZ','HH','WY','WU','TK','ZS','HK','WD','GW','MH','MF','YF','GC','PK','JP','VP','YG','GF','VC','HZ','WB','WC','DZ','FD','JN','XB','HV','CX','WM','JS','UW','KD','JC','KF','FK','TJ','WF','YK','BG','DK','QI','VM','BF','XD','JM','JK','MK','XS','KC','ZR','GJ','JH','KP','KV','JD','CJ','XG','XJ','ZD','GZ','YH','VH','VV','HG','YY','VG','MZ','WP','CW','PX','JJ','BK','JF','PZ','FJ','WG','JL','ZV','GV','MJ','VB','PW','BW','VW','JT','ZW','QL','FH','JB','VF','VK','QR','SX','XW','MQ','FW','FV','DX','KJ','XN','VQ','QO','JV','QM','QE','FX','QS','BZ','HQ','ZF','ZJ','PQ','QQ','XR','QB','HJ','GX','QC','YJ','YX','GQ','BQ','QT','JG','JW','QP','QN','BX','TQ','JY','QV','FQ','KX','QW','VX','VZ','FZ','XK','JX','KZ','QH','QD','WZ','VJ','WV','ZX','YQ','QF','JZ','WJ','HX','XZ','WX','WQ','JQ','QX','KQ','QY','QG','QZ','QK','QJ'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "bigram_frequencies = np.array([28642306,27599303,23987675,20108334,18188528,17889132,17471661,17137795,17135545,16053776,16043220,15999094,15611106,14776535,13881634,13403852,12839068,12788337,12469334,12252845,12162129,11767925,11702339,10797471,10629390,10479844,10366402,9921741,9824959,9806256,9755313,9704625,9597369,9486958,9350376,8948040,8899064,8532616,8489520,8264649,7761788,7454837,7314256,7158745,6982844,6916547,6869549,6831479,6501402,6391388,6379482,5959359,5867386,5713280,5703195,5671299,5499065,5432893,5428698,5260721,5255405,5241844,5209417,5192648,5153251,5025434,4905487,4575544,4535985,4487327,4271172,4257588,4232075,4171890,3925769,3881135,3698607,3573364,3473930,3321914,3275910,3232492,3219207,3216626,3190352,3131502,3063320,3051922,2993483,2911020,2854819,2853098,2765437,2749651,2727622,2722762,2699510,2694782,2622122,2555138,2534426,2489346,2412701,2381530,2352231,2293919,2286607,2248985,2228369,2216496,2169462,2096075,2082624,2053224,2036436,2010116,1982627,1927284,1908571,1907706,1900077,1806136,1789152,1755760,1752660,1667276,1634965,1616541,1610262,1577699,1560823,1547179,1537677,1524597,1509084,1508281,1504832,1459569,1458322,1455753,1421023,1406597,1388150,1329773,1308414,1302941,1280075,1269690,1253772,1253199,1246814,1241905,1212959,1209022,1206620,1185409,1181226,1165828,1140387,1107391,1098790,1098310,1095937,1093881,1079022,1072129,1066811,1020093,1018731,1017226,1015153,1014569,1008458,983685,961703,937006,912603,882550,860917,824133,820132,796875,790102,787382,784852,768894,758127,725594,723344,718973,699766,691623,668500,643725,641809,618917,618651,618602,578063,569051,540445,532762,527460,518357,513248,511995,498117,459829,421540,412761,384412,379311,377058,373468,355931,354392,353419,338775,330783,330010,329863,324846,316539,314965,310763,301316,297547,285120,279633,279558,272830,269252,240016,238792,234822,231240,210351,201316,200997,198839,193436,182266,175574,174475,173436,167332,161216,154455,152136,149314,145492,139196,137084,134881,133228,130549,130541,129843,126400,123342,117839,117167,115885,113582,113230,112679,112616,112238,111755,110089,101445,100184,99569,99228,92874,90401,88068,88065,85753,85590,83286,78319,77781,75985,74574,71957,70249,68579,65186,64486,62820,62619,61234,60326,58669,58448,56883,56704,54920,53366,52918,50963,48402,45540,45499,44827,43965,43768,43109,42568,41525,41141,38670,38492,37155,37038,35777,35461,35133,35011,34926,34907,34213,34171,33701,33537,33377,33375,33086,33014,32585,32534,32514,32194,30422,28281,28210,27443,26880,25768,25760,25018,24875,24672,24571,23668,23320,23226,23163,22813,21691,21260,20812,20210,20156,19469,19456,18914,18875,18500,17954,17951,17773,16849,16717,16621,16478,16454,15875,15769,15530,15430,15207,15050,14940,14827,14810,14403,14321,14320,14182,13801,13465,13328,12246,12089,11804,11730,11641,11534,11317,10977,10968,10908,10893,10803,10780,10599,10588,10461,10400,10301,10050,9836,9563,9472,9038,8919,8769,8685,8680,8489,8274,8257,8206,7903,7796,7733,7607,7400,7250,6804,6804,6524,6442,6440,6423,6420,6330,6324,6220,6000,5962,5947,5922,5833,5759,5741,5412,5404,5389,5274,5030,5021,4828,4599,4537,4489,4432,4260,4141,4111,4065,4009,3796,3786,3764,3719,3696,3648,3603,3546,3537,3360,3354,3279,3263,3263,3260,3244,3155,3067,2986,2978,2956,2954,2953,2950,2930,2919,2893,2873,2858,2758,2754,2747,2642,2596,2575,2542,2520,2407,2393,2388,2384,2260,2251,2231,2221,2215,2209,2205,2122,2100,2083,2018,1998,1993,1888,1883,1855,1851,1704,1687,1659,1642,1608,1545,1474,1471,1448,1431,1409,1398,1392,1356,1317,1308,1300,1251,1244,1242,1233,1172,1163,1161,1156,1144,1133,1132,1129,1089,1064,1060,1057,1052,1045,1026,997,957,956,942,922,918,893,883,863,856,850,849,841,841,806,784,782,774,748,742,737,732,691,691,678,647,647,620,599,592,579,568,563,563,532,532,511,475,471,465,463,443,439,426,421,420,419,415,407,391,366,366,355,353,337,324,324,313,308,296,291,281,266,261,257,251,244,231,227,222,221,219,212,207,191,186,177,176,169,168,167,165,154,143,140,140,139,137,135,126,123,118,113,104,101,100,92,90,90,90,90,88,83,81,79,73,72,68,68,62,56,55,48,45,38,33,32,25,23,20,19,18,15,14,11,7,6,4])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "46wIL5xzDzRS"
   },
   "source": [
    "## Speed matrix  <a name=\"speed\">\n",
    "### 24x24 relative Speed matrix between key pair (averaged for left/right symmetry)\n",
    "\n",
    "  - does not take into account order of key pairs (see Flow24x24 matrix)\n",
    "  - the original version was constructed with data from right-handed people\n",
    "  - 24 keys that don't require extending index or little fingers (\"home block keys\")\n",
    "\n",
    "### Home block keys\n",
    "\n",
    "        Left:            Right:\n",
    "     1  2  3  4       13 14 15 16 \n",
    "     5  6  7  8       17 18 19 20\n",
    "     9 10 11 12       21 22 23 24\n",
    "\n",
    "Interkey stroke times in milliseconds from Table 3 of <br>\n",
    "\"Estimation of digraph costs for keyboard layout optimization\", <br>\n",
    "A Iseri, Ma Eksioglu, International Journal of Industrial Ergonomics, 48, 127-138, 2015. <br>\n",
    "Key numbering in article and in spreadsheet:\n",
    "\n",
    "         Left:           Right:\n",
    "     1 4 7 10 13   16 19 22 25 28 31\n",
    "     2 5 8 11 14   17 20 23 26 29 32\n",
    "     3 6 9 12 15   18 21 24 27 30\n",
    "     \n",
    "### Load table of interkey speeds"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "095yG4iPDzRT"
   },
   "outputs": [],
   "source": [
    "# %load data/Time24x24.py\n",
    "# code/load_original_interkey_speeds.py\n",
    "#        Left:            Right:\n",
    "#    1  2  3  4 25   28 13 14 15 16 31 \n",
    "#    5  6  7  8 26   29 17 18 19 20 32\n",
    "#    9 10 11 12 27   30 21 22 23 24\n",
    "Time24x24 = np.array([\n",
    "[196,225,204,164,266,258,231,166,357,325,263,186,169,176,178,186,156,156,158,163,171,175,177,189],\n",
    "[225,181,182,147,239,245,196,150,289,296,229,167,162,169,170,178,148,148,150,155,163,167,169,182],\n",
    "[204,182,170,149,196,194,232,155,237,214,263,166,157,164,165,173,143,143,145,150,158,163,164,177],\n",
    "[164,147,149,169,160,161,157,226,165,185,234,257,154,162,163,171,141,141,143,148,156,160,162,175],\n",
    "[266,239,196,160,196,240,208,166,271,267,208,169,143,150,151,160,129,129,132,137,145,149,151,163],\n",
    "[258,245,194,161,240,181,183,149,245,256,184,150,138,145,146,154,124,124,126,131,139,144,145,158],\n",
    "[231,196,232,157,208,183,170,149,201,215,239,151,134,141,142,150,120,120,122,127,135,140,141,154],\n",
    "[166,150,155,226,166,149,149,169,160,147,170,221,133,140,141,150,119,119,122,126,135,139,141,153],\n",
    "[357,289,237,165,271,245,201,160,196,236,194,161,171,178,179,188,157,157,160,164,173,177,179,191],\n",
    "[325,296,214,185,267,256,215,147,236,181,184,157,166,173,174,182,152,152,154,159,167,172,173,186],\n",
    "[263,229,263,234,208,184,239,170,194,184,170,150,159,166,167,176,145,145,148,153,161,165,167,179],\n",
    "[186,167,166,257,169,150,151,221,161,157,150,169,153,160,161,169,139,139,141,146,154,159,160,173],\n",
    "[169,162,157,154,143,138,134,133,171,166,159,153,151,147,141,145,188,151,142,164,213,204,162,149],\n",
    "[176,169,164,162,150,145,141,140,178,173,166,160,147,151,189,209,137,207,191,206,149,227,208,197],\n",
    "[178,170,165,163,151,146,142,141,179,174,167,161,141,189,157,253,136,188,210,231,155,226,239,276],\n",
    "[186,178,173,171,160,154,150,150,188,182,176,169,145,209,253,170,147,206,251,233,164,268,362,271],\n",
    "[156,148,143,141,129,124,120,119,157,152,145,139,188,137,136,147,151,133,138,152,192,149,139,144],\n",
    "[156,148,143,141,129,124,120,119,157,152,145,139,151,207,188,206,133,151,179,183,145,204,183,201],\n",
    "[158,150,145,143,132,126,122,122,160,154,148,141,142,191,210,251,138,179,157,240,145,185,208,229],\n",
    "[163,155,150,148,137,131,127,126,164,159,153,146,164,206,231,233,152,183,240,170,160,220,293,242],\n",
    "[171,163,158,156,145,139,135,135,173,167,161,154,213,149,155,164,192,145,145,160,151,140,142,145],\n",
    "[175,167,163,160,149,144,140,139,177,172,165,159,204,227,226,268,149,204,185,220,140,151,175,188],\n",
    "[177,169,164,162,151,145,141,141,179,173,167,160,162,208,239,362,139,183,208,293,142,175,157,230],\n",
    "[189,182,177,175,163,158,154,153,191,186,179,173,149,197,276,271,144,201,229,242,145,188,230,170]])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# %load code/load_interkey_speeds24x24.py\n",
    "\n",
    "# Left/right symmetric version of the Time24x24 matrix\n",
    "# (The original version was constructed with data from right-handed people.)\n",
    "# A. Iseri, M. Eksioglu / International Journal of Industrial Ergonomics 48 (2015) 127e138\n",
    "\n",
    "#        Left:            Right:\n",
    "#     1  2  3  4       13 14 15 16 \n",
    "#     5  6  7  8       17 18 19 20\n",
    "#     9 10 11 12       21 22 23 24\n",
    "\n",
    "I = [ 1, 2, 3, 4,   5, 6, 7, 8,   9,10,11,12,  16,15,14,13,  20,19,18,17,  24,23,22,21]\n",
    "J = [16,15,14,13,  20,19,18,17,  24,23,22,21,   1, 2, 3, 4,   5, 6, 7, 8,   9,10,11,12]\n",
    "\n",
    "TimeSymmetric24x24 = np.ones((24,24))\n",
    "for i1, I1 in enumerate(I):\n",
    "    for i2, I2 in enumerate(I):\n",
    "        J1 = J[i1] - 1\n",
    "        J2 = J[i2] - 1\n",
    "        avgvalue = (Time24x24[I1-1,I2-1] + Time24x24[J1,J2]) / 2 \n",
    "        #print(Time24x24[I1-1,I2-1], Time24x24[J1,J2], avgvalue)\n",
    "        TimeSymmetric24x24[I1-1,I2-1] = avgvalue\n",
    "        TimeSymmetric24x24[J1,J2] = avgvalue\n",
    "\n",
    "# Normalize matrix with min-max scaling to a range with maximum = 1:\n",
    "newMin = np.min(Time24x24) / np.max(Time24x24)\n",
    "newMax = 1.0\n",
    "Time24x24 = newMin + (Time24x24 - np.min(Time24x24)) * (newMax - newMin) / (np.max(Time24x24) - np.min(Time24x24))\n",
    "\n",
    "# Convert relative interkey stroke times to relative speeds by subtracting from 1:\n",
    "Speed24x24 = 1 - Time24x24 + np.min(Time24x24)\n",
    "\n",
    "# Normalize matrix with min-max scaling to a range with maximum = 1:\n",
    "newMin = np.min(TimeSymmetric24x24) / np.max(TimeSymmetric24x24)\n",
    "newMax = 1.0\n",
    "TimeSymmetric24x24 = newMin + (TimeSymmetric24x24 - np.min(TimeSymmetric24x24)) * (newMax - newMin) / (np.max(TimeSymmetric24x24) - np.min(TimeSymmetric24x24))\n",
    "\n",
    "# Convert relative interkey stroke times to relative speeds by subtracting from 1:\n",
    "SpeedSymmetric24x24 = 1 - TimeSymmetric24x24 + np.min(TimeSymmetric24x24)\n",
    "\n",
    "# Print:\n",
    "#print_matrix_info(matrix_data=Speed24x24, matrix_label=\"Speed24x24\", nkeys=24, nlines=50)\n",
    "#heatmap(data=Speed24x24, title=\"Speed24x24\", xlabel=\"Key 1\", ylabel=\"Key 2\", print_output=print_output)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "tFfuA8zMDzRg"
   },
   "source": [
    "## Strength matrix  <a name=\"strength\">\n",
    "\n",
    "### Relative finger position STRENGTH matrix\n",
    "\n",
    "Finger strengths are based on peak keyboard reaction forces (in newtons) from Table 4 of <br>\n",
    "\"Keyboard Reaction Force and Finger Flexor Electromyograms during Computer Keyboard Work\" <br>\n",
    "BJ Martin, TJ Armstrong, JA Foulke, S Natarajan, Human Factors,1996,38(4),654-664:\n",
    "   \n",
    "    middle     2.36\n",
    "    index      2.26\n",
    "    ring       2.02\n",
    "    little     1.84\n",
    "    \n",
    "    index/middle:  0.9576271186440678\n",
    "    ring/middle:   0.8559322033898306\n",
    "    little/middle: 0.7796610169491526\n",
    "\n",
    "For reference, Table 1 of \"Ergonomic keyboard layout designed for the Filipino language\", 2016 (doi: 10.1007/978-3-319-41694-6_41) presents \"average finger strength of Filipinos [n=30, ages 16-36] measured in pounds\":\n",
    "   \n",
    "                L       R\n",
    "    little     3.77   4.27\n",
    "    ring       4.54   5.08\n",
    "    middle     5.65   6.37\n",
    "    index      6.09   6.57\n",
    "    \n",
    "    6.57/4.27 = 1.54\n",
    "    6.09/3.77 = 1.62\n",
    "    6.37/5.08 = 1.25\n",
    "    5.65/4.54 = 1.24\n",
    "    \n",
    "We won't use these results as I don't feel they represent relative strength relevant for typing: \"Respondents were asked to sit in upright position, with their wrists resting on a flat surface. A pinch gauge was placed within each finger's reach. The respondents were asked to exert maximum pressure on the device.\"\n",
    "    \n",
    "The following does not take into account order of key pairs (see Flow matrix).\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# %load code/load_strength_data.py\n",
    "# Normalize by the highest peak force (middle finger):\n",
    "middle_force = 2.36\n",
    "index_force = 2.26\n",
    "ring_force = 2.02\n",
    "little_force = 1.84\n",
    "middle_norm = 1.0\n",
    "index_norm = index_force / middle_force\n",
    "ring_norm = ring_force / middle_force\n",
    "little_norm = little_force / middle_force\n",
    "print('index/middle: {0}'.format(index_norm))\n",
    "print('ring/middle: {0}'.format(ring_norm))\n",
    "print('little/middle: {0}'.format(little_norm))\n",
    "\n",
    "# Relative left/right hand strength (assume equal):\n",
    "lf = 1.0\n",
    "rf = 1.0\n",
    "\n",
    "strengths24 = np.array((\n",
    "                    lf * little_norm, lf * ring_norm, lf * middle_norm, lf * index_norm,\n",
    "                    lf * little_norm, lf * ring_norm, lf * middle_norm, lf * index_norm,\n",
    "                    lf * little_norm, lf * ring_norm, lf * middle_norm, lf * index_norm,\n",
    "                    rf * index_norm, rf * middle_norm, rf * ring_norm, rf * little_norm,\n",
    "                    rf * index_norm, rf * middle_norm, rf * ring_norm, rf * little_norm,\n",
    "                    rf * index_norm, rf * middle_norm, rf * ring_norm, rf * little_norm))\n",
    "\n",
    "# Create a finger-pair position strength matrix by adding pairs of strength values:\n",
    "Strength24x24 = np.zeros((24, 24))\n",
    "for i in range(24):\n",
    "    Strength24x24[i,:] = strengths24\n",
    "Strength24x24 = (Strength24x24 + Strength24x24.transpose())\n",
    "\n",
    "# Normalize matrix with min-max scaling to a range with maximum = 1:\n",
    "#newMin = strength_factor\n",
    "newMin = min_strength_factor  # np.min(Strength24x24) / np.max(Strength24x24)\n",
    "newMax = 1.0\n",
    "Strength24x24 = newMin + (Strength24x24 - np.min(Strength24x24)) * (newMax - newMin) / (np.max(Strength24x24) - np.min(Strength24x24))\n",
    "\n",
    "# Print:\n",
    "print_matrix_info(matrix_data=Strength24x24, matrix_label=\"Strength24x24\", nkeys=24, nlines=10)\n",
    "heatmap(data=Strength24x24, title=\"Strength24x24\", xlabel=\"Key 1\", ylabel=\"Key 2\", print_output=print_output)\n",
    "\n",
    "# Save:\n",
    "if print_output:\n",
    "    file = open(\"Strength24x24.txt\", \"w+\")\n",
    "    file.write(str(Strength24x24))\n",
    "    file.close()\n",
    "\n",
    "\n",
    "penalty = 1.0  # Penalty for lateral (index, little) finger placement (1 = no penalty)\n",
    "\n",
    "strengths32 = np.array((lf * little_norm, lf * ring_norm, lf * middle_norm, lf * index_norm,\n",
    "                        lf * little_norm, lf * ring_norm, lf * middle_norm, lf * index_norm,\n",
    "                        lf * little_norm, lf * ring_norm, lf * middle_norm, lf * index_norm,\n",
    "                        rf * index_norm, rf * middle_norm, rf * ring_norm, rf * little_norm,\n",
    "                        rf * index_norm, rf * middle_norm, rf * ring_norm, rf * little_norm,\n",
    "                        rf * index_norm, rf * middle_norm, rf * ring_norm, rf * little_norm,\n",
    "                        lf * index_norm * penalty, lf * index_norm * penalty, lf * index_norm * penalty,\n",
    "                        rf * index_norm * penalty, rf * index_norm * penalty, rf * index_norm * penalty,\n",
    "                        rf * little_norm * penalty, rf * little_norm * penalty))\n",
    "\n",
    "# Create a finger-pair position strength matrix by adding pairs of strength values:\n",
    "Strength32x32 = np.zeros((32, 32))\n",
    "for i in range(32):\n",
    "    Strength32x32[i,:] = strengths32\n",
    "Strength32x32 = (Strength32x32 + Strength32x32.transpose())\n",
    "\n",
    "# Normalize matrix with min-max scaling to a range with maximum = 1:\n",
    "newMin = np.min(Strength32x32) / np.max(Strength32x32)\n",
    "newMax = 1.0\n",
    "Strength32x32 = newMin + (Strength32x32 - np.min(Strength32x32)) * (newMax - newMin) / (np.max(Strength32x32) - np.min(Strength32x32))\n",
    "\n",
    "# Print:\n",
    "print_matrix_info(matrix_data=Strength32x32, matrix_label=\"Strength32x32\", nkeys=32, nlines=10)\n",
    "heatmap(data=Strength32x32, title=\"Strength32x32\", xlabel=\"Key 1\", ylabel=\"Key 2\", print_output=print_output)\n",
    "\n",
    "# Save:\n",
    "if print_output:\n",
    "    file = open(\"Strength32x32.txt\", \"w+\")\n",
    "    file.write(str(Strength32x32))\n",
    "    file.close()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "Dnn5-8S5DzRy"
   },
   "source": [
    "## Flow matrix and Engram scoring model  <a name=\"flow\">\n",
    "\n",
    "The Flow24x24 matrix takes into account ease of transition between ordered pairs of keys.\n",
    "    \n",
    "Our optimization algorithm finds every permutation of a given set of letters, maps these letter permutations to a set of keys, and ranks these letter-key mappings according to a score reflecting ease of typing key pairs and frequency of letter pairs (bigrams). The score is the average of the scores for all possible bigrams in this arrangement. The score for each bigram is a product of the frequency of occurrence of that bigram, the frequency of each of the bigram’s characters, and flow, strength (and optional speed) factors for the key pair.\n",
    "\n",
    "#### Dvorak et al. (1936) defined eleven criteria for the design and evaluation of keyboard layouts:\n",
    "1.  Deviation from the balance of hand and finger loads should be as low as possible.\n",
    "2.  Percentage of tapping with the same fingers should be as low as possible.\n",
    "3.  Percentage of tapping that includes top row should be as low as possible.\n",
    "4.  Percentage of tapping that includes bottom row should be as low as possible.\n",
    "5.  Percentage of tapping in the home row should be as high as possible.\n",
    "6.  Percentage of tapping by alternating hands should be as high as possible.\n",
    "7.  Percentage of hurdles with the same finger should be as low as possible.\n",
    "8.  Percentage of hurdles with adjacent offset fingers should be as low as possible.\n",
    "9.  Percentage of hurdles with remote fingers should be as low as possible.\n",
    "10. Percentage of reach with the same finger should be as low as possible.\n",
    "11. Percentage of reach with adjacent offset fingers should be as low as possible.\n",
    "\n",
    "#### Synopsis of above criteria for pairwise key presses when touch typing:\n",
    "1. Alternate between hands.\n",
    "2. Balance finger loads, and avoid using the same finger.\n",
    "3. Avoid the upper and lower rows, and avoid skipping over the home row.\n",
    "4. Avoid tapping adjacent offset rows with the same or adjacent offset fingers.\n",
    "    \n",
    "### Factors to penalize strenuous key transitions\n",
    "\n",
    "Direction:\n",
    "    \n",
    "    - outward = 0.9: outward roll of fingers from the index to little finger (same hand)\n",
    "\n",
    "Dexterity:\n",
    "    \n",
    "    - side_above_3away = 0.9\n",
    "        - index and little finger type two keys, one or more rows apart (same hand)\n",
    "    - side_above_2away = 0.9^2 = 0.81\n",
    "        - index finger types key a row or two above ring finger key, or\n",
    "        - little finger types key a row or two above middle finger key (same hand)\n",
    "    - side_above_1away = 0.9^3 = 0.729\n",
    "        - index finger types key a row or two above middle finger key, or\n",
    "        - little finger types key a row or two above ring finger key (same hand)\n",
    "    - middle_above_ring = 0.9\n",
    "        - middle finger types key a row or two above ring finger key (same hand)\n",
    "    - ring_above_middle = 0.9^3 = 0.729\n",
    "        - ring finger types key a row or two above middle finger key (same hand)\n",
    "    - lateral = 0.9\n",
    "        - lateral movement of (index or little) finger outside of 8 vertical columns\n",
    "    \n",
    "Distance:\n",
    "    \n",
    "    - skip_row_3away = 0.9       \n",
    "        - index and little fingers type two keys that skip over home row (same hand)\n",
    "        - (e.g., one on bottom row, the other on top row)\n",
    "    - skip_row_2away = 0.9^3 = 0.729\n",
    "        - little and middle or index and ring fingers type two keys that skip over home row (same hand)\n",
    "    - skip_row_1away = 0.9^5 = 0.59049\n",
    "        - little and ring or middle and index fingers type two keys that skip over home row (same hand)\n",
    "\n",
    "Repetition:\n",
    " \n",
    "    - skip_row_0away = 0.9^4 = 0.6561\n",
    "        - same finger types two keys that skip over home row\n",
    "    - same_finger = 0.9^5 = 0.59049\n",
    "        - use same finger again for a different key\n",
    "        - cannot accompany outward, side_above, or adjacent_shorter_above \n",
    "\n",
    "Strength: Accounted for by the strength matrix (minimum value for the little finger = 0.9)    "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Example flow values for left side home block\n",
    "   \n",
    "No penalty (for same hand, both keys in the same row in an inward roll or repeating the same key):\n",
    "\n",
    "    2=>2, 2=>3, 3=>4, 2=>4, 1=>4\n",
    "\n",
    "    1  2  3  4\n",
    "    5  6  7  8\n",
    "    9 10 11 12\n",
    "\n",
    "Penalty = 0.9:\n",
    "\n",
    "    outward: 2=>1, 3=>1, 3=>2, 4=>1, 4=>2, 4=>3, 6=>5, 7=>6, 7=>5, 8=>7, 8=>6, 8=>5,... \n",
    "    middle_above_ring: 6=>3, 10=>7 \n",
    "    side_above_3away: 1=>8, 5=>4, 5=>12, 9=>8\n",
    "    index_above: 1=>4, 2=>4, 3=>4, 4=>4\n",
    "\n",
    "Penalty = 0.9^2:\n",
    "\n",
    "    middle_above_ring * outward: 3=>6, 7=>10\n",
    "    side_above_3away * outward: 8=>1, 4=>5, 12=>5, 8=>9\n",
    "    side_above_2away: 1=>7, 6=>4, 5=>11, 10=>8    \n",
    "    skip_row_3away * side_above_3away: 1=>12, 9=>4\n",
    "    skip_row_2away: 2=>12, 9=>3\n",
    "    ring_above_middle 2=>7, 6=>11\n",
    "    side_above_2away * outward: 7=>1, 4=>6, 11=>5, 8=>10\n",
    "    side_above_1away: 1=>6, 7=>4, 5=>10, 11=>8\n",
    "\n",
    "Penalty = 0.9^3:\n",
    "\n",
    "    skip_row_3away * side_above_3away * outward: 12=>1, 4=>9\n",
    "\n",
    "Penalty = 0.9^4:\n",
    "\n",
    "    ring_above_middle * outward: 7=>2, 11=>6\n",
    "    side_above_1away * outward: 4=>7, 6=>1, 10=>5, 4=>7\n",
    "\n",
    "Penalty = 0.9^5:\n",
    "\n",
    "    same_finger: 4=>8, 8=>4, 1=>5, 5=>1, 5=>9, 9=>5, 2=>6, 6=>2,...\n",
    "    skip_row_2away * side_above_2away: 10=>4, 1=>11\n",
    "    skip_row_1away: 1=>10, 9=>2, 3=>12\n",
    "\n",
    "Penalty = 0.9^6:\n",
    "\n",
    "    skip_row_2away * side_above_2away * outward: 4=>10, 11=>1\n",
    "    skip_row_1away * outward: 10=>1, 2=>9, 12=>3\n",
    "\n",
    "Penalty = 0.9^8\n",
    "\n",
    "    skip_row_1away * ring_above_middle: 2=>11\n",
    "    skip_row_1away * side_above_1away: 1=>10, 11=>4\n",
    "\n",
    "Penalty = 0.9^9\n",
    "\n",
    "    skip_row_1away * ring_above_middle * outward: 11=>2\n",
    "    skip_row_0away * same_finger: 1=>9, 9=>1, 4=>12, 12=>4, 2=>10, 10=>2, 3=>11, 11=>3     \n",
    "    skip_row_1away * side_above_1away * outward: 10=>1, 4=>11"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# %load code/load_flow_matrices.py\n",
    "# Penalizing factors for 24 keys  (1 = no penalty; set to less than 1 to penalize):\n",
    "\n",
    "# Dexterity\n",
    "side_above_3away = 0.9     # index and little finger type two keys, one or more rows apart (same hand)\n",
    "side_above_2away = 0.81    # index finger types key a row or two above ring finger key, or\n",
    "                           # little finger types key a row or two above middle finger key (same hand)\n",
    "side_above_1away = 0.729   # index finger types key a row or two above middle finger key, or\n",
    "                           # little finger types key a row or two above ring finger key (same hand)\n",
    "middle_above_ring = 0.9    # middle finger types key a row or two above ring finger key (same hand)\n",
    "ring_above_middle = 0.729  # ring finger types key a row or two above middle finger key (same hand)\n",
    "lateral = 0.9              # lateral movement of (index or little) finger outside of 8 vertical columns\n",
    "\n",
    "# Direction\n",
    "outward = 0.9              # outward roll of fingers from the index to little finger (same hand)\n",
    "\n",
    "# Distance\n",
    "skip_row_3away = 0.9       # index and little fingers type two keys that skip over home row (same hand)\n",
    "                           # (e.g., one on bottom row, the other on top row)\n",
    "skip_row_2away = 0.729     # little and middle or index and ring fingers type two keys that skip over home row (same hand)\n",
    "skip_row_1away = 0.59049   # little and ring or middle and index fingers type two keys that skip over home row (same hand)\n",
    "\n",
    "# Repetition\n",
    "skip_row_0away = 0.6561    # same finger types two keys that skip over home row\n",
    "same_finger = 0.59049      # use same finger again for a different key\n",
    "\n",
    "\n",
    "# Unused or redundant parameters\n",
    "same_hand = 1.0            # (addressed by splitting up the most frequent letters across left/right sides above)\n",
    "not_home_row = 1.0         # at least one key not on home row\n",
    "side_top = 1.0             # index or little finger types top corner key\n",
    "shorter_above = 1.0        # (taken care of by side_above_[1,2,3]away parameters)\n",
    "adjacent_offset = 1.0      # (taken care of by side_above_1away, middle_above_ring, ring_above_middle parameters)\n",
    "inside_top = 1.0           # index finger types top corner key (taken care of by side_above_1away parameter)\n",
    "index_above = 1.0          # index finger types top corner key (unless other bigram key is in the top row for the same hand)\n",
    "                           # (taken care of by side_above_[1,2,3]away parameters)\n",
    "\n",
    "\n",
    "def create_24x24_flow_matrix(not_home_row, side_top, side_above_3away, side_above_2away, side_above_1away, \n",
    "                             middle_above_ring, ring_above_middle, outward, skip_row_3away, \n",
    "                             skip_row_2away, skip_row_1away, skip_row_0away, same_finger, lateral, \n",
    "                             same_hand, shorter_above, adjacent_offset, inside_top, index_above):\n",
    "\n",
    "    all_24_keys = [1,2,3,4, 5,6,7,8, 9,10,11,12, 13,14,15,16, 17,18,19,20, 21,22,23,24]\n",
    "\n",
    "    # Create a matrix and multiply by flow factors that promote easy interkey transitions:\n",
    "    T = np.ones((24, 24))\n",
    "\n",
    "    # 7.  Promote alternating between hands over uncomfortable transitions with the same hand.\n",
    "    if same_hand < 1.0:\n",
    "    #    1  2  3  4   13 14 15 16  \n",
    "    #    5  6  7  8   17 18 19 20 \n",
    "    #    9 10 11 12   21 22 23 24\n",
    "        for i in range(0,12):\n",
    "            for j in range(0,12):\n",
    "                T[i,j] *= same_hand\n",
    "        for i in range(12,24):\n",
    "            for j in range(12,24):\n",
    "                T[i,j] *= same_hand\n",
    "\n",
    "    # 8.  Promote little-to-index-finger roll-ins over index-to-little-finger outwards.\n",
    "    #    1  2  3  4   13 14 15 16  \n",
    "    #    5  6  7  8   17 18 19 20 \n",
    "    #    9 10 11 12   21 22 23 24\n",
    "    if outward < 1.0:\n",
    "\n",
    "        # same-row roll-outs:\n",
    "        roll_ins = [[1,2],[2,3],[3,4], [5,6],[6,7],[7,8], [9,10],[10,11],[11,12],\n",
    "                    [16,15],[15,14],[14,13], [20,19],[19,18],[18,17], [24,23],[23,22],[22,21]]\n",
    "        for x in roll_ins:\n",
    "            T[x[1]-1, x[0]-1] *= outward\n",
    "\n",
    "        # same-row roll-outs, skipping keys:\n",
    "        roll_ins_skip_keys = [[1,3],[2,4],[1,4], [5,7],[6,8],[5,8], [9,11],[10,12],[9,12],\n",
    "                              [16,14],[15,13],[16,13], [20,18],[19,17],[20,17], [24,22],[23,21],[24,21]]\n",
    "        for x in roll_ins_skip_keys:\n",
    "            T[x[1]-1, x[0]-1] *= outward\n",
    "\n",
    "        # adjacent-row roll-outs:\n",
    "        roll_ins_adj_rows = [[1,6],[1,7],[1,8],[2,7],[2,8],[3,8], [5,2],[5,3],[5,4],[6,3],[6,4],[7,4],\n",
    "                             [5,10],[5,11],[5,12],[6,11],[6,12],[7,12], [9,6],[9,7],[9,8],[10,7],[10,8],[11,8],\n",
    "                             [16,19],[16,18],[16,17],[15,18],[15,17],[14,17], [20,15],[20,14],[20,13],[19,14],[19,13],[18,13],\n",
    "                             [20,23],[20,22],[20,21],[19,22],[19,21],[18,21], [24,19],[24,18],[24,17],[23,18],[23,17],[22,17]]\n",
    "        for x in roll_ins_adj_rows:\n",
    "            T[x[1]-1, x[0]-1] *= outward\n",
    "\n",
    "        # upper<->lower row roll-outs:\n",
    "        roll_ins_skip_home = [[1,10],[1,11],[1,12],[2,11],[2,12],[3,12], [9,2],[9,3],[9,4],[10,3],[10,4],[11,4],\n",
    "                              [16,23],[16,22],[16,21],[15,22],[15,21],[14,21], [24,15],[24,14],[24,13],[23,14],[23,13],[22,13]]\n",
    "        for x in roll_ins_skip_home:\n",
    "            T[x[1]-1, x[0]-1] *= outward\n",
    "\n",
    "    # 9.  Avoid stretching shorter fingers up and longer fingers down.\n",
    "    #    1  2  3  4   13 14 15 16  \n",
    "    #    5  6  7  8   17 18 19 20 \n",
    "    #    9 10 11 12   21 22 23 24\n",
    "    if index_above < 1.0:\n",
    "        for x in [4]:\n",
    "            for y in [4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24]:\n",
    "                T[x-1, y-1] *= index_above\n",
    "                T[y-1, x-1] *= index_above\n",
    "        for x in [13]:\n",
    "            for y in [1,2,3,4,5,6,7,8,9,10,11,12,13,17,18,19,20,21,22,23,24]:\n",
    "                T[x-1, y-1] *= index_above\n",
    "                T[y-1, x-1] *= index_above\n",
    "    if inside_top < 1.0:\n",
    "        for x in [4,13]:\n",
    "            for j in range(0,24):\n",
    "                T[x-1, j] *= inside_top\n",
    "                T[j, x-1] *= inside_top\n",
    "    if side_top < 1.0:\n",
    "        for x in [1,4,13,16]:\n",
    "            for j in range(0,24):\n",
    "                T[x-1, j] *= side_top\n",
    "                T[j, x-1] *= side_top\n",
    "    if side_above_1away < 1.0:\n",
    "        for x in [1]:\n",
    "            for y in [6,10]:\n",
    "                T[x-1, y-1] *= side_above_1away\n",
    "                T[y-1, x-1] *= side_above_1away\n",
    "        for x in [5]:\n",
    "            for y in [10]:\n",
    "                T[x-1, y-1] *= side_above_1away\n",
    "                T[y-1, x-1] *= side_above_1away\n",
    "        for x in [4]:\n",
    "            for y in [7,11]:\n",
    "                T[x-1, y-1] *= side_above_1away\n",
    "                T[y-1, x-1] *= side_above_1away\n",
    "        for x in [8]:\n",
    "            for y in [11]:\n",
    "                T[x-1, y-1] *= side_above_1away\n",
    "                T[y-1, x-1] *= side_above_1away\n",
    "        for x in [13]:\n",
    "            for y in [18,22]:\n",
    "                T[x-1, y-1] *= side_above_1away\n",
    "                T[y-1, x-1] *= side_above_1away\n",
    "        for x in [17]:\n",
    "            for y in [22]:\n",
    "                T[x-1, y-1] *= side_above_1away\n",
    "                T[y-1, x-1] *= side_above_1away\n",
    "        for x in [16]:\n",
    "            for y in [19,23]:\n",
    "                T[x-1, y-1] *= side_above_1away\n",
    "                T[y-1, x-1] *= side_above_1away\n",
    "        for x in [20]:\n",
    "            for y in [23]:\n",
    "                T[x-1, y-1] *= side_above_1away\n",
    "                T[y-1, x-1] *= side_above_1away\n",
    "    if side_above_2away < 1.0:\n",
    "        for x in [1]:\n",
    "            for y in [7,11]:\n",
    "                T[x-1, y-1] *= side_above_2away\n",
    "                T[y-1, x-1] *= side_above_2away\n",
    "        for x in [5]:\n",
    "            for y in [11]:\n",
    "                T[x-1, y-1] *= side_above_2away\n",
    "                T[y-1, x-1] *= side_above_2away\n",
    "        for x in [4]:\n",
    "            for y in [6,10]:\n",
    "                T[x-1, y-1] *= side_above_2away\n",
    "                T[y-1, x-1] *= side_above_2away\n",
    "        for x in [8]:\n",
    "            for y in [10]:\n",
    "                T[x-1, y-1] *= side_above_2away\n",
    "                T[y-1, x-1] *= side_above_2away\n",
    "        for x in [13]:\n",
    "            for y in [19,23]:\n",
    "                T[x-1, y-1] *= side_above_2away\n",
    "                T[y-1, x-1] *= side_above_2away\n",
    "        for x in [17]:\n",
    "            for y in [23]:\n",
    "                T[x-1, y-1] *= side_above_2away\n",
    "                T[y-1, x-1] *= side_above_2away\n",
    "        for x in [16]:\n",
    "            for y in [18,22]:\n",
    "                T[x-1, y-1] *= side_above_2away\n",
    "                T[y-1, x-1] *= side_above_2away\n",
    "        for x in [20]:\n",
    "            for y in [22]:\n",
    "                T[x-1, y-1] *= side_above_2away\n",
    "                T[y-1, x-1] *= side_above_2away\n",
    "    if side_above_3away < 1.0:\n",
    "        for x in [1]:\n",
    "            for y in [8,12]:\n",
    "                T[x-1, y-1] *= side_above_3away\n",
    "                T[y-1, x-1] *= side_above_3away\n",
    "        for x in [5]:\n",
    "            for y in [12]:\n",
    "                T[x-1, y-1] *= side_above_3away\n",
    "                T[y-1, x-1] *= side_above_3away\n",
    "        for x in [4]:\n",
    "            for y in [5,9]:\n",
    "                T[x-1, y-1] *= side_above_3away\n",
    "                T[y-1, x-1] *= side_above_3away\n",
    "        for x in [8]:\n",
    "            for y in [9]:\n",
    "                T[x-1, y-1] *= side_above_3away\n",
    "                T[y-1, x-1] *= side_above_3away\n",
    "        for x in [13]:\n",
    "            for y in [20,24]:\n",
    "                T[x-1, y-1] *= side_above_3away\n",
    "                T[y-1, x-1] *= side_above_3away\n",
    "        for x in [17]:\n",
    "            for y in [24]:\n",
    "                T[x-1, y-1] *= side_above_3away\n",
    "                T[y-1, x-1] *= side_above_3away\n",
    "        for x in [16]:\n",
    "            for y in [17,21]:\n",
    "                T[x-1, y-1] *= side_above_3away\n",
    "                T[y-1, x-1] *= side_above_3away\n",
    "        for x in [20]:\n",
    "            for y in [21]:\n",
    "                T[x-1, y-1] *= side_above_3away\n",
    "                T[y-1, x-1] *= side_above_3away\n",
    "    if shorter_above < 1.0:\n",
    "        for x in [1]:\n",
    "            for y in [6,7,8,10,11,12]:\n",
    "                T[x-1, y-1] *= shorter_above\n",
    "                T[y-1, x-1] *= shorter_above\n",
    "        for x in [2]:\n",
    "            for y in [7,11]:\n",
    "                T[x-1, y-1] *= shorter_above\n",
    "                T[y-1, x-1] *= shorter_above\n",
    "        for x in [4]:\n",
    "            for y in [6,7,10,11]:\n",
    "                T[x-1, y-1] *= shorter_above\n",
    "                T[y-1, x-1] *= shorter_above\n",
    "\n",
    "        for x in [5]:\n",
    "            for y in [10,11,12]:\n",
    "                T[x-1, y-1] *= shorter_above\n",
    "                T[y-1, x-1] *= shorter_above\n",
    "        for x in [6]:\n",
    "            for y in [11]:\n",
    "                T[x-1, y-1] *= shorter_above\n",
    "                T[y-1, x-1] *= shorter_above\n",
    "        for x in [8]:\n",
    "            for y in [10,11]:\n",
    "                T[x-1, y-1] *= shorter_above\n",
    "                T[y-1, x-1] *= shorter_above\n",
    "\n",
    "        for x in [16]:\n",
    "            for y in [17,18,19,21,22,23]:\n",
    "                T[x-1, y-1] *= shorter_above\n",
    "                T[y-1, x-1] *= shorter_above\n",
    "        for x in [15]:\n",
    "            for y in [18,22]:\n",
    "                T[x-1, y-1] *= shorter_above\n",
    "                T[y-1, x-1] *= shorter_above\n",
    "        for x in [13]:\n",
    "            for y in [18,19,22,23]:\n",
    "                T[x-1, y-1] *= shorter_above\n",
    "                T[y-1, x-1] *= shorter_above\n",
    "\n",
    "        for x in [20]:\n",
    "            for y in [21,22,23]:\n",
    "                T[x-1, y-1] *= shorter_above\n",
    "                T[y-1, x-1] *= shorter_above\n",
    "        for x in [19]:\n",
    "            for y in [22]:\n",
    "                T[x-1, y-1] *= shorter_above\n",
    "                T[y-1, x-1] *= shorter_above\n",
    "        for x in [17]:\n",
    "            for y in [22,23]:\n",
    "                T[x-1, y-1] *= shorter_above\n",
    "                T[y-1, x-1] *= shorter_above\n",
    "\n",
    "    if ring_above_middle < 1.0:\n",
    "        ring_above_middles =  [[2,7],[6,11],[2,11],\n",
    "                            [15,18],[19,22],[15,22]]\n",
    "        for x in ring_above_middles:\n",
    "            T[x[0]-1, x[1]-1] *= ring_above_middle\n",
    "            T[x[1]-1, x[0]-1] *= ring_above_middle\n",
    "\n",
    "    if middle_above_ring < 1.0:\n",
    "        middle_above_rings =  [[6,3],[10,7],[10,3],\n",
    "                            [19,14],[23,18],[23,14]]\n",
    "        for x in middle_above_rings:\n",
    "            T[x[0]-1, x[1]-1] *= middle_above_ring\n",
    "            T[x[1]-1, x[0]-1] *= middle_above_ring\n",
    "\n",
    "    # 10. Avoid using the same finger.\n",
    "    #    1  2  3  4   13 14 15 16  \n",
    "    #    5  6  7  8   17 18 19 20 \n",
    "    #    9 10 11 12   21 22 23 24\n",
    "    if same_finger < 1.0:\n",
    "        same_fingers = [[1,5],[5,9],[1,9], [2,6],[6,10],[2,10], [3,7],[7,11],[3,11], [4,8],[8,12],[4,12],\n",
    "                        [13,17],[17,21],[13,21], [14,18],[18,22],[14,22], [15,19],[19,23],[15,23], [16,20],[20,24],[16,24]] \n",
    "        for x in same_fingers:\n",
    "            T[x[0]-1, x[1]-1] *= same_finger\n",
    "            T[x[1]-1, x[0]-1] *= same_finger\n",
    "\n",
    "    # 11. Avoid the upper and lower rows.\n",
    "    #    1  2  3  4   13 14 15 16  \n",
    "    #    5  6  7  8   17 18 19 20 \n",
    "    #    9 10 11 12   21 22 23 24\n",
    "    if not_home_row < 1.0:\n",
    "        not_home_row_keys = [1,2,3,4, 9,10,11,12, 13,14,15,16, 21,22,23,24]\n",
    "        for x in not_home_row_keys:\n",
    "            for j in range(0,23):\n",
    "                T[x-1, j] *= not_home_row\n",
    "                T[j, x-1] *= not_home_row\n",
    "\n",
    "    # 12. Avoid skipping over the home row.\n",
    "    #    1  2  3  4   13 14 15 16  \n",
    "    #    5  6  7  8   17 18 19 20 \n",
    "    #    9 10 11 12   21 22 23 24\n",
    "    if skip_row_0away < 1.0:\n",
    "        skip_top = [1, 2, 3, 4, 13,14,15,16] \n",
    "        skip_bot = [9,10,11,12, 21,22,23,24] \n",
    "        for ix, x in enumerate(skip_top):\n",
    "            y = skip_bot[ix]\n",
    "            T[x-1, y-1] *= skip_row_0away\n",
    "            T[y-1, x-1] *= skip_row_0away\n",
    "    if skip_row_1away < 1.0:\n",
    "        skip_top = [1, 2, 2, 3, 3, 4, 13,14,14,15,15,16] \n",
    "        skip_bot = [10,9,11,10,12,11, 22,21,23,22,24,23] \n",
    "        for ix, x in enumerate(skip_top):\n",
    "            y = skip_bot[ix]\n",
    "            T[x-1, y-1] *= skip_row_1away\n",
    "            T[y-1, x-1] *= skip_row_1away\n",
    "    if skip_row_2away < 1.0:\n",
    "        skip_top = [1,  2,3, 4, 13,14,15,16] \n",
    "        skip_bot = [11,12,9,10, 23,24,21,22] \n",
    "        for ix, x in enumerate(skip_top):\n",
    "            y = skip_bot[ix]\n",
    "            T[x-1, y-1] *= skip_row_2away\n",
    "            T[y-1, x-1] *= skip_row_2away\n",
    "    if skip_row_3away < 1.0:\n",
    "        skip_top = [1, 4, 13,16] \n",
    "        skip_bot = [12,9, 24,21] \n",
    "        for ix, x in enumerate(skip_top):\n",
    "            y = skip_bot[ix]\n",
    "            T[x-1, y-1] *= skip_row_3away\n",
    "            T[y-1, x-1] *= skip_row_3away\n",
    "\n",
    "    Flow24x24 = T\n",
    "\n",
    "    # Normalize matrix with min-max scaling to a range with maximum = 1:\n",
    "    newMin = np.min(Flow24x24) / np.max(Flow24x24)\n",
    "    newMax = 1.0\n",
    "    Flow24x24 = newMin + (Flow24x24 - np.min(Flow24x24)) * (newMax - newMin) / (np.max(Flow24x24) - np.min(Flow24x24))\n",
    "\n",
    "    return Flow24x24\n",
    "\n",
    "Flow24x24 = create_24x24_flow_matrix(not_home_row, side_top, \n",
    "    side_above_3away, side_above_2away, side_above_1away, middle_above_ring, ring_above_middle, outward, \n",
    "    skip_row_3away, skip_row_2away, skip_row_1away, skip_row_0away, same_finger, lateral, same_hand, \n",
    "    shorter_above, adjacent_offset, inside_top, index_above)\n",
    "\n",
    "# Print:\n",
    "print_matrix_info(matrix_data=Flow24x24, matrix_label=\"Flow24x24\", nkeys=24, nlines=30)\n",
    "heatmap(data=Flow24x24, title=\"Flow24x24\", xlabel=\"Key 1\", ylabel=\"Key 2\", print_output=print_output)\n",
    "\n",
    "\n",
    "def create_32x32_flow_matrix(not_home_row, side_top, side_above_3away, side_above_2away, side_above_1away, \n",
    "                             middle_above_ring, ring_above_middle, outward, skip_row_3away, \n",
    "                             skip_row_2away, skip_row_1away, skip_row_0away, same_finger, lateral, \n",
    "                             same_hand, shorter_above, adjacent_offset, inside_top, index_above):\n",
    "\n",
    "    all_32_keys = [1,2,3,4, 5,6,7,8, 9,10,11,12, 13,14,15,16, 17,18,19,20, 21,22,23,24, \n",
    "                   25,26,27, 28,29,30, 31,32]\n",
    "\n",
    "    # Create a matrix and multiply by flow factors that promote easy interkey transitions:\n",
    "    T = np.ones((32, 32))\n",
    "\n",
    "    if lateral < 1.0:\n",
    "        for x in all_32_keys:\n",
    "            for y in [25,26,27, 28,29,30, 31,32]:\n",
    "                T[x-1, y-1] *= lateral\n",
    "                T[y-1, x-1] *= lateral    \n",
    "\n",
    "    # 7.  Promote alternating between hands over uncomfortable transitions with the same hand.\n",
    "    if same_hand < 1.0:\n",
    "        for i in [1,2,3,4,5,6,7,8,9,10,11,12, 25,26,27]:\n",
    "            for j in [1,2,3,4,5,6,7,8,9,10,11,12, 25,26,27]:\n",
    "                T[i-1,j-1] *= same_hand\n",
    "        for i in [13,14,15,16,17,18,19,20,21,22,23,24, 28,29,30,31,32]:\n",
    "            for j in [13,14,15,16,17,18,19,20,21,22,23,24, 28,29,30,31,32]:\n",
    "                T[i-1,j-1] *= same_hand\n",
    "\n",
    "    # 8.  Promote little-to-index-finger roll-ins over index-to-little-finger outsward rolls.\n",
    "    # Penalize (index, little) finger lateral movements:\n",
    "    #  1  2  3  4 25   28 13 14 15 16 31 \n",
    "    #  5  6  7  8 26   29 17 18 19 20 32\n",
    "    #  9 10 11 12 27   30 21 22 23 24\n",
    "    if outward < 1.0:\n",
    "\n",
    "        # same-row roll-outs:\n",
    "        roll_ins = [[1,2],[2,3],[3,4], [5,6],[6,7],[7,8], [9,10],[10,11],[11,12],\n",
    "                    [16,15],[15,14],[14,13], [20,19],[19,18],[18,17], [24,23],[23,22],[22,21]]\n",
    "        for x in roll_ins:\n",
    "            T[x[1]-1, x[0]-1] *= outward\n",
    "\n",
    "        # same-row roll-outs, skipping keys:\n",
    "        roll_ins_skip_keys = [[1,3],[2,4],[1,4], [5,7],[6,8],[5,8], [9,11],[10,12],[9,12],\n",
    "                              [16,14],[15,13],[16,13], [20,18],[19,17],[20,17], [24,22],[23,21],[24,21]]\n",
    "                              #[1,25],[2,25],[3,25],\n",
    "                              #[5,26],[6,26],[7,26],\n",
    "                              #[9,27],[10,27],[11,27],\n",
    "                              #[16,28],[15,28],[14,28],\n",
    "                              #[20,29],[19,29],[18,29],\n",
    "                              #[24,30],[23,30],[22,30],\n",
    "                              #[31,15],[31,14],[31,13],[31,28],\n",
    "                              #[32,19],[32,18],[32,17],[32,29]]\n",
    "        for x in roll_ins_skip_keys:\n",
    "            T[x[1]-1, x[0]-1] *= outward\n",
    "\n",
    "        # adjacent-row roll-outs:\n",
    "        #  1  2  3  4 25   28 13 14 15 16 31 \n",
    "        #  5  6  7  8 26   29 17 18 19 20 32\n",
    "        #  9 10 11 12 27   30 21 22 23 24\n",
    "        roll_ins_adj_rows = [[1,6],[1,7],[1,8],[2,7],[2,8],[3,8], \n",
    "                             [5,2],[5,3],[5,4],[6,3],[6,4],[7,4],\n",
    "                             [5,10],[5,11],[5,12],[6,11],[6,12],[7,12], \n",
    "                             [9,6],[9,7],[9,8],[10,7],[10,8],[11,8],\n",
    "                             [16,19],[16,18],[16,17],[15,18],[15,17],[14,17], \n",
    "                             [20,15],[20,14],[20,13],[19,14],[19,13],[18,13],\n",
    "                             [20,23],[20,22],[20,21],[19,22],[19,21],[18,21], \n",
    "                             [24,19],[24,18],[24,17],[23,18],[23,17],[22,17]]\n",
    "                             #[5,25],[6,25],[7,25],[8,25],\n",
    "                             #[5,27],[6,27],[7,27],[8,27],\n",
    "                             #[1,26],[2,26],[3,26],[4,26],\n",
    "                             #[9,26],[10,26],[11,26],[12,26],\n",
    "                             #[16,29],[15,29],[14,29],[13,29],\n",
    "                             #[24,29],[23,29],[22,29],[21,29],\n",
    "                             #[20,28],[19,28],[18,28],[17,28],\n",
    "                             #[20,30],[19,30],[18,30],[17,30],\n",
    "                             #[31,20],[31,19],[31,18],[31,17],[31,29],\n",
    "                             #[32,16],[32,15],[32,14],[32,13],[32,28],\n",
    "                             #[32,24],[32,23],[32,22],[32,21],[32,30]]\n",
    "        for x in roll_ins_adj_rows:\n",
    "            T[x[1]-1, x[0]-1] *= outward\n",
    "\n",
    "        # upper<->lower row roll-outs:\n",
    "        roll_ins_skip_home = [[1,10],[1,11],[1,12],[2,11],[2,12],[3,12],\n",
    "                              [9,2],[9,3],[9,4],[10,3],[10,4],[11,4],\n",
    "                              [16,23],[16,22],[16,21],[15,22],[15,21],[14,21],\n",
    "                              [24,15],[24,14],[24,13],[23,14],[23,13],[22,13]]\n",
    "                              #[16,30],[15,30],[14,30],[13,30],\n",
    "                              #[9,25],[10,25],[11,25],[12,25],\n",
    "                              #[24,28],[23,28],[22,28],[21,28],\n",
    "                              #[1,27],[2,27],[3,27],[4,27], \n",
    "                              #[31,24],[31,23],[31,22],[31,21],[31,30]]\n",
    "        for x in roll_ins_skip_home:\n",
    "            T[x[1]-1, x[0]-1] *= outward\n",
    "\n",
    "    # 9.  Avoid stretching shorter fingers up and longer fingers down.\n",
    "    #  1  2  3  4 25   28 13 14 15 16 31 \n",
    "    #  5  6  7  8 26   29 17 18 19 20 32\n",
    "    #  9 10 11 12 27   30 21 22 23 24\n",
    "    if index_above < 1.0:\n",
    "        for x in [4]:\n",
    "            for y in [4,5,6,7,8,26,9,10,11,12,27,28,13,14,15,16,31,29,17,18,19,20,32,30,21,22,23,24]:\n",
    "                T[x-1, y-1] *= index_above\n",
    "                T[y-1, x-1] *= index_above\n",
    "        for x in [25]:\n",
    "            for y in [25,5,6,7,8,26,9,10,11,12,27,28,13,14,15,16,31,29,17,18,19,20,32,30,21,22,23,24]:\n",
    "                T[x-1, y-1] *= index_above\n",
    "                T[y-1, x-1] *= index_above\n",
    "        for x in [13]:\n",
    "            for y in [1,2,3,4,25,5,6,7,8,26,9,10,11,12,27,13,29,17,18,19,20,32,30,21,22,23,24]:\n",
    "                T[x-1, y-1] *= index_above\n",
    "                T[y-1, x-1] *= index_above\n",
    "        for x in [28]:\n",
    "            for y in [1,2,3,4,25,5,6,7,8,26,9,10,11,12,27,28,29,17,18,19,20,32,30,21,22,23,24]:\n",
    "                T[x-1, y-1] *= index_above\n",
    "                T[y-1, x-1] *= index_above\n",
    "    if inside_top < 1.0:\n",
    "        for x in [4,25,28,13]:\n",
    "            for j in range(0,32):\n",
    "                T[x-1, j] *= inside_top\n",
    "                T[j, x-1] *= inside_top\n",
    "    if side_top < 1.0:\n",
    "        for x in [1,4,25,28,13,16,31]:\n",
    "            for j in range(0,32):\n",
    "                T[x-1, j] *= side_top\n",
    "                T[j, x-1] *= side_top\n",
    "    if side_above_1away < 1.0:\n",
    "        for x in [1]:\n",
    "            for y in [6,10]:\n",
    "                T[x-1, y-1] *= side_above_1away\n",
    "                T[y-1, x-1] *= side_above_1away\n",
    "        for x in [5]:\n",
    "            for y in [10]:\n",
    "                T[x-1, y-1] *= side_above_1away\n",
    "                T[y-1, x-1] *= side_above_1away\n",
    "        for x in [4,25]:\n",
    "            for y in [7,11]:\n",
    "                T[x-1, y-1] *= side_above_1away\n",
    "                T[y-1, x-1] *= side_above_1away\n",
    "        for x in [8,26]:\n",
    "            for y in [11]:\n",
    "                T[x-1, y-1] *= side_above_1away\n",
    "                T[y-1, x-1] *= side_above_1away\n",
    "        for x in [13,28]:\n",
    "            for y in [18,22]:\n",
    "                T[x-1, y-1] *= side_above_1away\n",
    "                T[y-1, x-1] *= side_above_1away\n",
    "        for x in [17,29]:\n",
    "            for y in [22]:\n",
    "                T[x-1, y-1] *= side_above_1away\n",
    "                T[y-1, x-1] *= side_above_1away\n",
    "        for x in [16,31]:\n",
    "            for y in [19,23]:\n",
    "                T[x-1, y-1] *= side_above_1away\n",
    "                T[y-1, x-1] *= side_above_1away\n",
    "        for x in [20,32]:\n",
    "            for y in [23]:\n",
    "                T[x-1, y-1] *= side_above_1away\n",
    "                T[y-1, x-1] *= side_above_1away\n",
    "    if side_above_2away < 1.0:\n",
    "        for x in [1]:\n",
    "            for y in [7,11]:\n",
    "                T[x-1, y-1] *= side_above_2away\n",
    "                T[y-1, x-1] *= side_above_2away\n",
    "        for x in [5]:\n",
    "            for y in [11]:\n",
    "                T[x-1, y-1] *= side_above_2away\n",
    "                T[y-1, x-1] *= side_above_2away\n",
    "        for x in [4,25]:\n",
    "            for y in [6,10]:\n",
    "                T[x-1, y-1] *= side_above_2away\n",
    "                T[y-1, x-1] *= side_above_2away\n",
    "        for x in [8,26]:\n",
    "            for y in [10]:\n",
    "                T[x-1, y-1] *= side_above_2away\n",
    "                T[y-1, x-1] *= side_above_2away\n",
    "        for x in [13,28]:\n",
    "            for y in [19,23]:\n",
    "                T[x-1, y-1] *= side_above_2away\n",
    "                T[y-1, x-1] *= side_above_2away\n",
    "        for x in [17,29]:\n",
    "            for y in [23]:\n",
    "                T[x-1, y-1] *= side_above_2away\n",
    "                T[y-1, x-1] *= side_above_2away\n",
    "        for x in [16,31]:\n",
    "            for y in [18,22]:\n",
    "                T[x-1, y-1] *= side_above_2away\n",
    "                T[y-1, x-1] *= side_above_2away\n",
    "        for x in [20,32]:\n",
    "            for y in [22]:\n",
    "                T[x-1, y-1] *= side_above_2away\n",
    "                T[y-1, x-1] *= side_above_2away\n",
    "    if side_above_3away < 1.0:\n",
    "        for x in [1]:\n",
    "            for y in [8,12,26,27]:\n",
    "                T[x-1, y-1] *= side_above_3away\n",
    "                T[y-1, x-1] *= side_above_3away\n",
    "        for x in [5]:\n",
    "            for y in [12,27]:\n",
    "                T[x-1, y-1] *= side_above_3away\n",
    "                T[y-1, x-1] *= side_above_3away\n",
    "        for x in [4,25]:\n",
    "            for y in [5,9]:\n",
    "                T[x-1, y-1] *= side_above_3away\n",
    "                T[y-1, x-1] *= side_above_3away\n",
    "        for x in [8,26]:\n",
    "            for y in [9]:\n",
    "                T[x-1, y-1] *= side_above_3away\n",
    "                T[y-1, x-1] *= side_above_3away\n",
    "        for x in [13,28]:\n",
    "            for y in [20,24,32]:\n",
    "                T[x-1, y-1] *= side_above_3away\n",
    "                T[y-1, x-1] *= side_above_3away\n",
    "        for x in [17,29]:\n",
    "            for y in [24]:\n",
    "                T[x-1, y-1] *= side_above_3away\n",
    "                T[y-1, x-1] *= side_above_3away\n",
    "        for x in [16,31]:\n",
    "            for y in [17,21,29,30]:\n",
    "                T[x-1, y-1] *= side_above_3away\n",
    "                T[y-1, x-1] *= side_above_3away\n",
    "        for x in [20,32]:\n",
    "            for y in [21,30]:\n",
    "                T[x-1, y-1] *= side_above_3away\n",
    "                T[y-1, x-1] *= side_above_3away\n",
    "\n",
    "\n",
    "    #  1  2  3  4 25   28 13 14 15 16 31 \n",
    "    #  5  6  7  8 26   29 17 18 19 20 32\n",
    "    #  9 10 11 12 27   30 21 22 23 24\n",
    "    if shorter_above < 1.0:\n",
    "        for x in [1]:\n",
    "            for y in [6,7,8,26,10,11,12,27]:\n",
    "                T[x-1, y-1] *= shorter_above\n",
    "                T[y-1, x-1] *= shorter_above\n",
    "        for x in [2]:\n",
    "            for y in [7,11]:\n",
    "                T[x-1, y-1] *= shorter_above\n",
    "                T[y-1, x-1] *= shorter_above\n",
    "        for x in [4]:\n",
    "            for y in [6,7,10,11]:\n",
    "                T[x-1, y-1] *= shorter_above\n",
    "                T[y-1, x-1] *= shorter_above\n",
    "        for x in [25]:\n",
    "            for y in [6,7,10,11]:\n",
    "                T[x-1, y-1] *= shorter_above\n",
    "                T[y-1, x-1] *= shorter_above\n",
    "\n",
    "        for x in [5]:\n",
    "            for y in [10,11,12,27]:\n",
    "                T[x-1, y-1] *= shorter_above\n",
    "                T[y-1, x-1] *= shorter_above\n",
    "        for x in [6]:\n",
    "            for y in [11]:\n",
    "                T[x-1, y-1] *= shorter_above\n",
    "                T[y-1, x-1] *= shorter_above\n",
    "        for x in [8]:\n",
    "            for y in [10,11]:\n",
    "                T[x-1, y-1] *= shorter_above\n",
    "                T[y-1, x-1] *= shorter_above\n",
    "        for x in [26]:\n",
    "            for y in [10,11]:\n",
    "                T[x-1, y-1] *= shorter_above\n",
    "                T[y-1, x-1] *= shorter_above\n",
    "\n",
    "        for x in [16]:\n",
    "            for y in [29,17,18,19,30,21,22,23]:\n",
    "                T[x-1, y-1] *= shorter_above\n",
    "                T[y-1, x-1] *= shorter_above\n",
    "        for x in [31]:\n",
    "            for y in [29,17,18,19,30,21,22,23]:\n",
    "                T[x-1, y-1] *= shorter_above\n",
    "                T[y-1, x-1] *= shorter_above\n",
    "        for x in [15]:\n",
    "            for y in [18,22]:\n",
    "                T[x-1, y-1] *= shorter_above\n",
    "                T[y-1, x-1] *= shorter_above\n",
    "        for x in [13]:\n",
    "            for y in [18,19,22,23]:\n",
    "                T[x-1, y-1] *= shorter_above\n",
    "                T[y-1, x-1] *= shorter_above\n",
    "        for x in [28]:\n",
    "            for y in [18,19,22,23]:\n",
    "                T[x-1, y-1] *= shorter_above\n",
    "                T[y-1, x-1] *= shorter_above\n",
    "\n",
    "        for x in [20]:\n",
    "            for y in [30,21,22,23]:\n",
    "                T[x-1, y-1] *= shorter_above\n",
    "                T[y-1, x-1] *= shorter_above\n",
    "        for x in [32]:\n",
    "            for y in [30,21,22,23]:\n",
    "                T[x-1, y-1] *= shorter_above\n",
    "                T[y-1, x-1] *= shorter_above\n",
    "        for x in [19]:\n",
    "            for y in [22]:\n",
    "                T[x-1, y-1] *= shorter_above\n",
    "                T[y-1, x-1] *= shorter_above\n",
    "        for x in [17]:\n",
    "            for y in [22,23]:\n",
    "                T[x-1, y-1] *= shorter_above\n",
    "                T[y-1, x-1] *= shorter_above\n",
    "        for x in [29]:\n",
    "            for y in [22,23]:\n",
    "                T[x-1, y-1] *= shorter_above\n",
    "                T[y-1, x-1] *= shorter_above\n",
    "\n",
    "    if ring_above_middle < 1.0:\n",
    "        ring_above_middles =  [[2,7],[6,11],[2,11],\n",
    "                            [15,18],[19,22],[15,22]]\n",
    "        for x in ring_above_middles:\n",
    "            T[x[0]-1, x[1]-1] *= ring_above_middle\n",
    "            T[x[1]-1, x[0]-1] *= ring_above_middle\n",
    "\n",
    "    if middle_above_ring < 1.0:\n",
    "        middle_above_rings =  [[6,3],[10,7],[10,3],\n",
    "                            [19,14],[23,18],[23,14]]\n",
    "        for x in middle_above_rings:\n",
    "            T[x[0]-1, x[1]-1] *= middle_above_ring\n",
    "            T[x[1]-1, x[0]-1] *= middle_above_ring\n",
    "\n",
    "    # 10. Avoid using the same finger.\n",
    "    #  1  2  3  4 25   28 13 14 15 16 31 \n",
    "    #  5  6  7  8 26   29 17 18 19 20 32\n",
    "    #  9 10 11 12 27   30 21 22 23 24\n",
    "    if same_finger < 1.0:\n",
    "        same_fingers = [[1,5],[5,9],[1,9], [2,6],[6,10],[2,10], \n",
    "                        [3,7],[7,11],[3,11], [4,8],[8,12],[4,12],\n",
    "                        [25,26],[26,27],[25,27], [28,29],[29,30],[28,30], [31,32],\n",
    "                        [4,25],[4,26],[4,27], [8,25],[8,26],[8,27], [12,25],[12,26],[12,27],\n",
    "                        [13,28],[13,29],[13,30], [17,28],[17,29],[17,30], [21,28],[21,29],[21,30],\n",
    "                        [31,16],[31,20],[31,24], [32,16],[32,20],[32,24],\n",
    "                        [13,17],[17,21],[13,21], [14,18],[18,22],[14,22], \n",
    "                        [15,19],[19,23],[15,23], [16,20],[20,24],[16,24]] \n",
    "        for x in same_fingers:\n",
    "            T[x[0]-1, x[1]-1] *= same_finger\n",
    "            T[x[1]-1, x[0]-1] *= same_finger\n",
    "\n",
    "    # 11. Avoid the upper and lower rows.\n",
    "    if not_home_row < 1.0:\n",
    "        not_home_row_keys = [1,2,3,4,25, 9,10,11,12,27, 28,13,14,15,16,31, 30,21,22,23,24]\n",
    "        for x in not_home_row_keys:\n",
    "            for j in range(0,32):\n",
    "                T[x-1, j] *= not_home_row\n",
    "                T[j, x-1] *= not_home_row\n",
    "                \n",
    "    # 12. Avoid skipping over the home row.\n",
    "    #  1  2  3  4 25   28 13 14 15 16 31 \n",
    "    #  5  6  7  8 26   29 17 18 19 20 32\n",
    "    #  9 10 11 12 27   30 21 22 23 24\n",
    "    if skip_row_0away < 1.0:\n",
    "        skip_top = [1, 2, 3, 4, 4,25,25, 28,28,13,13,14,15,16,31] \n",
    "        skip_bot = [9,10,11,12,27,12,27, 30,21,30,21,22,23,24,24] \n",
    "        for ix, x in enumerate(skip_top):\n",
    "            y = skip_bot[ix]\n",
    "            T[x-1, y-1] *= skip_row_0away\n",
    "            T[y-1, x-1] *= skip_row_0away\n",
    "    if skip_row_1away < 1.0:\n",
    "        skip_top = [1, 2, 2, 3, 3, 4, 4,25, 28,13,13,14,14,15,15,16,31] \n",
    "        skip_bot = [10,9,11,10,12,11,27,11, 22,30,22,21,23,22,24,23,23] \n",
    "        for ix, x in enumerate(skip_top):\n",
    "            y = skip_bot[ix]\n",
    "            T[x-1, y-1] *= skip_row_1away\n",
    "            T[y-1, x-1] *= skip_row_1away\n",
    "    if skip_row_2away < 1.0:\n",
    "        skip_top = [1,  2,3, 4,25, 28,13,14,15,16,31] \n",
    "        skip_bot = [11,12,9,10,10, 23,23,24,21,22,22] \n",
    "        for ix, x in enumerate(skip_top):\n",
    "            y = skip_bot[ix]\n",
    "            T[x-1, y-1] *= skip_row_2away\n",
    "            T[y-1, x-1] *= skip_row_2away\n",
    "    if skip_row_3away < 1.0:\n",
    "        skip_top = [1, 4,25, 28,13,16,16,31,31] \n",
    "        skip_bot = [12,9, 9, 24,24,21,30,21,30] \n",
    "        for ix, x in enumerate(skip_top):\n",
    "            y = skip_bot[ix]\n",
    "            T[x-1, y-1] *= skip_row_3away\n",
    "            T[y-1, x-1] *= skip_row_3away\n",
    "                \n",
    "    Flow32x32 = T\n",
    "\n",
    "    # Normalize matrix with min-max scaling to a range with maximum = 1:\n",
    "    newMin = np.min(Flow32x32) / np.max(Flow32x32)\n",
    "    newMax = 1.0\n",
    "    Flow32x32 = newMin + (Flow32x32 - np.min(Flow32x32)) * (newMax - newMin) / (np.max(Flow32x32) - np.min(Flow32x32))\n",
    "\n",
    "    return Flow32x32\n",
    "\n",
    "Flow32x32 = create_32x32_flow_matrix(not_home_row, side_top, \n",
    "    side_above_3away, side_above_2away, side_above_1away, middle_above_ring, ring_above_middle, outward, \n",
    "    skip_row_3away, skip_row_2away, skip_row_1away, skip_row_0away, same_finger, lateral, same_hand, \n",
    "    shorter_above, adjacent_offset, inside_top, index_above)\n",
    "\n",
    "# Print:\n",
    "print_matrix_info(matrix_data=Flow32x32, matrix_label=\"Flow32x32\", nkeys=32, nlines=30)\n",
    "heatmap(data=Flow32x32, title=\"Flow32x32\", xlabel=\"Key 1\", ylabel=\"Key 2\", print_output=print_output)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "WMvP493uDzSU"
   },
   "source": [
    "## Combine Strength and Flow matrices  <a name=\"strengthflow\">"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 695
    },
    "colab_type": "code",
    "id": "UP7FUBR2DzSX",
    "outputId": "5dc11788-2c69-4f69-ab60-a07ac17e092f"
   },
   "outputs": [],
   "source": [
    "# %load code/combine_scoring_matrices.py\n",
    "# 24 keys:\n",
    "Factors24x24 = Flow24x24\n",
    "if apply_strength:\n",
    "    Factors24x24 = Strength24x24 * Factors24x24\n",
    "\n",
    "# Print:\n",
    "print_matrix_info(matrix_data=Factors24x24, matrix_label=\"Factors24x24\", nkeys=24, nlines=30)\n",
    "heatmap(data=Factors24x24, title=\"Factors24x24\", xlabel=\"Key 1\", ylabel=\"Key 2\", print_output=print_output)\n",
    "\n",
    "# Save:\n",
    "if print_output:\n",
    "    file = open(\"Factors24x24.txt\", \"w+\")\n",
    "    file.write(str(Factors24x24))\n",
    "    file.close()\n",
    "\n",
    "\n",
    "# 32 keys:\n",
    "Factors32x32 = Flow32x32\n",
    "if apply_strength:\n",
    "    Factors32x32 = Strength32x32 * Factors32x32\n",
    "\n",
    "# Print:\n",
    "print_matrix_info(matrix_data=Factors32x32, matrix_label=\"Factors32x32\", nkeys=32, nlines=30)\n",
    "heatmap(data=Factors32x32, title=\"Factors32x32\", xlabel=\"Key 1\", ylabel=\"Key 2\", print_output=print_output)\n",
    "\n",
    "# Save:\n",
    "if print_output:\n",
    "    file = open(\"Factors32x32.txt\", \"w+\")\n",
    "    file.write(str(Factors32x32))\n",
    "    file.close()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Four steps\n",
    "\n",
    "We will assign letters to keys by choosing the arrangement with the highest score according to our scoring model. However, there are over four hundred septillion, or four hundred trillion trillion (26! = 403,291,461,126,605,635,584,000,000, or 4.032914611 E+26) possible arrangements of 26 letters (24! = 6.204484017 E+23), so we will arrange the letters in four steps, based on ergonomics principles. These consist of (Step 1) assigning the eight most frequent letters to different keys, optimizing assignment of the remaining (Step 2) eight most frequent letters, and (Step 3) eight least frequent letters (besides Z and Q), and (Step 4) exchanging letters. \n",
    "\n",
    "## Step 1: Define the shape of the key layout to minimize lateral finger movements<a name=\"step1\">\n",
    "\n",
    "We will assign 24 letters to 8 columns of keys separated by two middle columns reserved for punctuation. These 8 columns require no lateral finger movements when touch typing, since there is one column per finger. The most comfortable keys include the left and right home rows (keys 5-8 and 17-20), the top-center keys (2,3 and 14,15) that allow the longer middle and ring fingers to uncurl upwards, as well as the bottom corner keys (9,12 and 21,24) that allow the shorter fingers to curl downwards. We will assign the two least frequent letters, W and K, to the two hardest-to-reach keys lying outside the 24-key columns in the upper right (25 and 26):\n",
    "\n",
    "        Left:            Right:\n",
    "     1  2  3  4       13 14 15 16 25\n",
    "     5  6  7  8       17 18 19 20 26\n",
    "     9 10 11 12       21 22 23 24\n",
    "\n",
    "We will consider the most comfortable keys to be those typed by either hand on the home row, by the ring and middle finger above the home row, and by the index and little finger below the home row, with a preference for the strongest (index and middle) fingers:\n",
    "    \n",
    "     -  2  3  -        - 14 15  -  \n",
    "     5  6  7  8       17 18 19 20  \n",
    "     9  -  - 12       21  -  - 24"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "REInHU9tdYLP"
   },
   "source": [
    "## Step 2: Arrange the most frequent letters based on comfort and bigram frequencies  <a name=\"step2\">\n",
    "\n",
    "Below, we will arrange vowels on one side and the most frequent consonants to the other side to encourage balance and alternation across hands. Since aside from the letters W and K there is symmetry across left and right sides, we will decide later which side the vowels and which side the most frequent consonants should go.\n",
    "\n",
    "### Vowels\n",
    "    \n",
    "**E, A, O**, S, N, **I**, R, L, D, C, T, U, M, P, B, G, V, Q, Y, F, H, J, Z, X, K, W\n",
    "                         \n",
    "Bigrams that contain two non-repeating vowels, and their frequencies:\n",
    "  \n",
    "    UE\t17135545\n",
    "    IO\t10629390\n",
    "    IA\t 9704625\n",
    "    IE \t 8264649\n",
    "    UA\t 2727622\n",
    "    UI\t 2699510\n",
    "    EA   1634965\n",
    "    AU   1206620\n",
    "    EO\t  787382\n",
    "    AI\t  768894\n",
    "    EI\t  578063\n",
    "    EU\t  338775\n",
    "    IU\t  314965\n",
    "    AE\t  279633\n",
    "    OI\t  193436\n",
    "    OE\t  182266\n",
    "    UO\t  175574\n",
    "    OU\t  174475\n",
    "    OA\t  117839\n",
    "    AO\t   88065\n",
    "  \n",
    "We will assign the most frequent vowels (E=294897235, A=271738665, O=201996963, I=151438547) to four of the six most comfortable keys on the left side of the keyboard (keys 2,3,5,6,7,8). We will assign the letter E, the most frequent in the Spanish language, to the strongest (middle) finger on the home row, and assign the other three vowels such that (1) the home row keys typed by the index and middle fingers are not left vacant, and any top-frequency bigram (more than 1 million instances) (2) does not use the same finger and (3) reads from left to right (ex: UE, not EU) for ease of typing (inward roll from little to index finger vs. outward roll from index to little finger). These constraints lead to 4 arrangements of the 4 vowels:\n",
    "\n",
    "    - - O -    - - O -    - O - -    - - - -    \n",
    "    - I E A    I - E A    I - E A    I O E A\n",
    "    - - - -    - - - -    - - - -    - - - -\n",
    "\n",
    "### Consonants\n",
    "\n",
    "E, A, O, **S, N**, I, **R, L, D, C, T**, U, M, P, B, G, V, Q, Y, F, H, J, Z, X, K, W\n",
    "\n",
    "On the right side of the keyboard, we will assign four of the seven most frequent consonants (S=162205879, N=161440601, R=147218050, L=124544026, D=113430944, C=99562807, T=99294129) to the four home row keys. We will assign the letter S, the most frequent consonant in the Spanish language, to the strongest (middle) finger on the home row.\n",
    "    \n",
    "Bigrams that contain two non-repeating consonants, with frequencies greater than 1 million:\n",
    "\n",
    "    NT\t13403852\n",
    "    ST\t10479844\n",
    "    TR   6982844\n",
    "    ND\t 6501402\n",
    "    PR\t 5241844\n",
    "    NC\t 4535985\n",
    "    RT\t 3473930\n",
    "    BR\t 3216626\n",
    "    MP\t 3063320\n",
    "    CH\t 2911020\n",
    "    SP\t 2412701\n",
    "    NS\t 2352231\n",
    "    MB\t 2169462\n",
    "    GR\t 2082624\n",
    "    BL\t 2053224\n",
    "    CT\t 1927284\n",
    "    RM\t 1908571\n",
    "    SC\t 1907706\n",
    "    CR\t 1537677\n",
    "    RD\t 1524597\n",
    "    RS\t 1509084\n",
    "    PL\t 1459569\n",
    "    RC\t 1280075\n",
    "    NG\t 1253772\n",
    "    LT\t 1253199\n",
    "    DR\t 1140387\n",
    "    RN\t 1072129\n",
    "    RG\t 1017226\n",
    "\n",
    "Bigrams from the above list that contain two of the most frequent consonants (S, N, R, L, D, C, T):\n",
    "    \n",
    "    NT\t13403852\n",
    "    ST\t10479844\n",
    "    TR   6982844\n",
    "    ND\t 6501402\n",
    "    NC\t 4535985\n",
    "    RT\t 3473930\n",
    "    NS\t 2352231\n",
    "    CT\t 1927284\n",
    "    SC\t 1907706\n",
    "    CR\t 1537677\n",
    "    RD\t 1524597\n",
    "    RS\t 1509084\n",
    "    RC\t 1280075\n",
    "    LT\t 1253199\n",
    "    DR\t 1140387\n",
    "    RN\t 1072129\n",
    "\n",
    "As with the left side, letters are placed so that top-frequency bigrams read from right to left (ex: \"NT\" read as TN, not NT) except when both sequences have more than 1 million instances (TR/RT, CR/RC, and RD/DR), for ease of typing. The above constraints lead to 4 arrangements of the consonants:\n",
    "\n",
    "    - - - -    - - - -    - - - -    - - - -\n",
    "    L S N R    D S N R    C S N R    T S N R\n",
    "    - - - -    - - - -    - - - -    - - - -\n",
    "\n",
    "The resulting 16 initial layouts, each with 16 unassigned keys, are represented below with the three rows on the left and right side of the keyboard as a linear string of letters, with unassigned keys denoted by “-”.\n",
    "    \n",
    "    Pattern 1:\n",
    "    \n",
    "    --O- -IEA ----    ---- LSNR ----\n",
    "    --O- -IEA ----    ---- DSNR ----\n",
    "    --O- -IEA ----    ---- CSNR ----\n",
    "    --O- -IEA ----    ---- TSNR ----\n",
    "\n",
    "    Pattern 2:\n",
    "    \n",
    "    --O- I-EA ----    ---- LSNR ----\n",
    "    --O- I-EA ----    ---- DSNR ----\n",
    "    --O- I-EA ----    ---- CSNR ----\n",
    "    --O- I-EA ----    ---- TSNR ----\n",
    "\n",
    "    Pattern 3:\n",
    "    \n",
    "    -O-- I-EA ----    ---- LSNR ----\n",
    "    -O-- I-EA ----    ---- DSNR ----\n",
    "    -O-- I-EA ----    ---- CSNR ----\n",
    "    -O-- I-EA ----    ---- TSNR ----\n",
    "\n",
    "    Pattern 4:\n",
    "    \n",
    "    ---- IOEA ----    ---- LSNR ----\n",
    "    ---- IOEA ----    ---- DSNR ----\n",
    "    ---- IOEA ----    ---- CSNR ----\n",
    "    ---- IOEA ----    ---- TSNR ----    \n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Step 3: Optimize assignment of the remaining letters <a name=\"step3\">\n",
    "    \n",
    "We want to assign letters to the 16 unassigned keys in each of the above 16 layouts based on our scoring model. That would mean scoring all possible arrangements for each layout and choosing the arrangement with the highest score, but since there are over 20 trillion (16!) possible ways of arranging 16 letters, we will break up the assignment into two stages for the most frequent and least frequent remaining letters. \n",
    "    \n",
    "### Most frequent letters\n",
    "We will compute scores for every possible arrangement of the seven most frequent of the remaining letters (in bold below) assigned to vacancies among the most comfortable sixteen keys.\n",
    "\n",
    "E, A, O, S, N, I, R, **L, D, C, T, U, P, M, B, G**, V, Q, H, F, Y, J, Z, X, K, W\n",
    "\n",
    "        Left:            Right:\n",
    "     -  2  3  -        - 14 15  -\n",
    "     5  6  7  8       17 18 19 20\n",
    "     9  -  - 12       21  -  - 24\n",
    "\n",
    "Since there are 40,320 (8!) possible combinations of eight letters for each of the 16 layouts, we need to score and evaluate 645,120 layouts. To score each arrangement of letters, we construct a frequency matrix where we multiply a matrix containing the frequency of each ordered pair of letters (bigram) by our flow and strength matrices to compute a score.\n",
    "    \n",
    "### Least frequent letters\n",
    "Next we will compute scores for every possible (40,320 = 8!) arrangement of the least frequent eight letters (in bold below, besides W and K) in the remaining keys, after substituting in the 16 top results of the above, for another 645,120 layouts:\n",
    "\n",
    "E, A, O, S, N, I, R, L, D, C, T, U, P, M, B, G, **V, Q, H, F, Y, J, Z, X**, K, W\n",
    "\n",
    "        Left:            Right:\n",
    "     1  -  -  4       13  -  - 16\n",
    "     -  -  -  -        -  -  -  -\n",
    "     - 10 11  -        - 22 23  -\n",
    "     \n",
    "### Further optimize layouts by exchanging more letters\n",
    "\n",
    "If we relax the above fixed initializations and permit further exchange of letters, then we can search for even higher-scoring layouts. As a final optimization step we exchange letters, eight keys at a time (8! = 40,320) selected twice in 14 different ways, in each of the above 16 layouts, to score a total of over 18 million combinations. We allow the following keys to exchange letters:\n",
    "\n",
    "    1. Top rows\n",
    "    2. Bottom rows\n",
    "    3. Top and bottom rows on the right side\n",
    "    4. Top and bottom rows on the left side\n",
    "    5. Top right and bottom left rows\n",
    "    6. Top left and bottom right rows\n",
    "    7. Center of the top and bottom rows on both sides\n",
    "    8. The eight corners\n",
    "    9. Left half of the top and bottom rows on both sides\n",
    "    10. Right half of the top and bottom rows on both sides\n",
    "    11. Left half of non-home rows on the left and right half of the same rows on the right\n",
    "    12. Right half of non-home rows on the left and left half of the same rows on the right\n",
    "    13. Top center and lower sides\n",
    "    14. Top sides and lower center\n",
    "    15. Repeat 1-14"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "\"\"\"\n",
    "NOTE: This procedure takes hours to run.\n",
    "\n",
    "    Pattern 1:\n",
    "    \n",
    "    --O- -IEA ----    ---- LSNR ----\n",
    "    --O- -IEA ----    ---- DSNR ----\n",
    "    --O- -IEA ----    ---- CSNR ----\n",
    "    --O- -IEA ----    ---- TSNR ----\n",
    "\n",
    "    Pattern 2:\n",
    "    \n",
    "    --O- I-EA ----    ---- LSNR ----\n",
    "    --O- I-EA ----    ---- DSNR ----\n",
    "    --O- I-EA ----    ---- CSNR ----\n",
    "    --O- I-EA ----    ---- TSNR ----\n",
    "\n",
    "    Pattern 3:\n",
    "    \n",
    "    -O-- I-EA ----    ---- LSNR ----\n",
    "    -O-- I-EA ----    ---- DSNR ----\n",
    "    -O-- I-EA ----    ---- CSNR ----\n",
    "    -O-- I-EA ----    ---- TSNR ----\n",
    "\n",
    "    Pattern 4:\n",
    "    \n",
    "    ---- IOEA ----    ---- LSNR ----\n",
    "    ---- IOEA ----    ---- DSNR ----\n",
    "    ---- IOEA ----    ---- CSNR ----\n",
    "    ---- IOEA ----    ---- TSNR ----    \n",
    "    \n",
    "\"\"\"\n",
    "fixed_letter_lists1 = [\n",
    "    ['O','I','E','A','L','S','N','R'],\n",
    "    ['O','I','E','A','D','S','N','R'],\n",
    "    ['O','I','E','A','C','S','N','R'],\n",
    "    ['O','I','E','A','T','S','N','R'],\n",
    "    ['O','I','E','A','L','S','N','R'],\n",
    "    ['O','I','E','A','D','S','N','R'],\n",
    "    ['O','I','E','A','C','S','N','R'],\n",
    "    ['O','I','E','A','T','S','N','R'],\n",
    "    ['O','I','E','A','L','S','N','R'],\n",
    "    ['O','I','E','A','D','S','N','R'],\n",
    "    ['O','I','E','A','C','S','N','R'],\n",
    "    ['O','I','E','A','T','S','N','R'],\n",
    "    ['I','O','E','A','L','S','N','R'],\n",
    "    ['I','O','E','A','D','S','N','R'],\n",
    "    ['I','O','E','A','C','S','N','R'],\n",
    "    ['I','O','E','A','T','S','N','R']]\n",
    "\n",
    "# Keys for step 1:\n",
    "#     -  2  3  -        - 14 15  -\n",
    "#     5  6  7  8       17 18 19 20\n",
    "#     9  -  - 12       21  -  - 24\n",
    "keys1  = [2,3,   5,6,7,8, 9,12,  14,15, 17,18,19,20, 21,24]\n",
    "\n",
    "# Indices for step 1:\n",
    "#     -  0  1  -        -  8  9  -\n",
    "#     2  3  4  5       10 11 12 13\n",
    "#     6  -  -  7       14  -  - 15\n",
    "fixed_letter_index_lists1 = [[1,   3,4,5, 10,11,12,13],\n",
    "                             [1,   3,4,5, 10,11,12,13],\n",
    "                             [1,   3,4,5, 10,11,12,13],\n",
    "                             [1,   3,4,5, 10,11,12,13],\n",
    "                             [1, 2,  4,5, 10,11,12,13],\n",
    "                             [1, 2,  4,5, 10,11,12,13],\n",
    "                             [1, 2,  4,5, 10,11,12,13],\n",
    "                             [1, 2,  4,5, 10,11,12,13],\n",
    "                             [0, 2,  4,5, 10,11,12,13],\n",
    "                             [0, 2,  4,5, 10,11,12,13],\n",
    "                             [0, 2,  4,5, 10,11,12,13],\n",
    "                             [0, 2,  4,5, 10,11,12,13],\n",
    "                             [   2,3,4,5, 10,11,12,13],\n",
    "                             [   2,3,4,5, 10,11,12,13],\n",
    "                             [   2,3,4,5, 10,11,12,13],\n",
    "                             [   2,3,4,5, 10,11,12,13]]\n",
    "open_letter_index_lists1  = [[0, 2,   6,7, 8,9, 14,15],\n",
    "                             [0, 2,   6,7, 8,9, 14,15],\n",
    "                             [0, 2,   6,7, 8,9, 14,15],\n",
    "                             [0, 2,   6,7, 8,9, 14,15],\n",
    "                             [0,   3, 6,7, 8,9, 14,15],\n",
    "                             [0,   3, 6,7, 8,9, 14,15],\n",
    "                             [0,   3, 6,7, 8,9, 14,15],\n",
    "                             [0,   3, 6,7, 8,9, 14,15],\n",
    "                             [1,   3, 6,7, 8,9, 14,15],\n",
    "                             [1,   3, 6,7, 8,9, 14,15],\n",
    "                             [1,   3, 6,7, 8,9, 14,15],\n",
    "                             [1,   3, 6,7, 8,9, 14,15],\n",
    "                             [0,1,    6,7, 8,9, 14,15],\n",
    "                             [0,1,    6,7, 8,9, 14,15],\n",
    "                             [0,1,    6,7, 8,9, 14,15],\n",
    "                             [0,1,    6,7, 8,9, 14,15]]\n",
    "\n",
    "# All 24 key indices:\n",
    "#     0  1  2  3       12 13 14 15\n",
    "#     4  5  6  7       16 17 18 19\n",
    "#     8  9 10 11       20 21 22 23\n",
    "# Open indices:\n",
    "#     0  -  -  3       12  -  - 15\n",
    "#     -  -  -  -        -  -  -  -\n",
    "#     -  9 10  -        - 21 22  -\n",
    "fixed_letter_indices2 = [1,2, 4,5,6,7, 8,11, 13,14, 16,17,18,19, 20,23]\n",
    "open_letter_indices2  = [0,3, 9,10, 12,15, 21,22]\n",
    "fixed_letter_index_lists3 = [[2,   5,6,7, 16,17,18,19],\n",
    "                             [2,   5,6,7, 16,17,18,19],\n",
    "                             [2,   5,6,7, 16,17,18,19],\n",
    "                             [2,   5,6,7, 16,17,18,19],\n",
    "                             [2, 4,  6,7, 16,17,18,19],\n",
    "                             [2, 4,  6,7, 16,17,18,19],\n",
    "                             [2, 4,  6,7, 16,17,18,19],\n",
    "                             [2, 4,  6,7, 16,17,18,19],\n",
    "                             [1, 4,  6,7, 16,17,18,19],\n",
    "                             [1, 4,  6,7, 16,17,18,19],\n",
    "                             [1, 4,  6,7, 16,17,18,19],\n",
    "                             [1, 4,  6,7, 16,17,18,19],\n",
    "                             [   4,5,6,7, 16,17,18,19],\n",
    "                             [   4,5,6,7, 16,17,18,19],\n",
    "                             [   4,5,6,7, 16,17,18,19],\n",
    "                             [   4,5,6,7, 16,17,18,19]]\n",
    "\n",
    "# Loop through initialized layouts with assigned vowels and consonants \n",
    "top_layouts = []\n",
    "nlists = len(fixed_letter_lists1)\n",
    "for ilist, fixed_letters1 in enumerate(fixed_letter_lists1):\n",
    "  if ilist > 10:\n",
    "    fixed_letter_indices1 = fixed_letter_index_lists1[ilist]\n",
    "    fixed_letter_indices3 = fixed_letter_index_lists3[ilist]\n",
    "    open_letter_indices1 = open_letter_index_lists1[ilist]\n",
    "\n",
    "    print('Layout {0}'.format(ilist+1))\n",
    "    print(*fixed_letters1)\n",
    "\n",
    "    print(\"Most frequent letters\")\n",
    "    top_permutation1, top_score1, letter_permutations1 = permute_optimize_keys(fixed_letters1, fixed_letter_indices1, \n",
    "                                                         open_letter_indices1, letters24, keys1, Factors24x24, \n",
    "                                                         bigrams, bigram_frequencies, min_score=0, verbose=False)\n",
    "    fixed_letters2 = top_permutation1\n",
    "    \n",
    "    print(\"Least frequent remaining letters\")\n",
    "    top_permutation2, top_score2, letter_permutations2 = permute_optimize_keys(fixed_letters2, fixed_letter_indices2, \n",
    "                                                         open_letter_indices2, letters24, keys24, Factors24x24, \n",
    "                                                         bigrams, bigram_frequencies, min_score=0, verbose=False)\n",
    "    fixed_letters3 = top_permutation2\n",
    "\n",
    "    print(\"Further optimize layouts by exchanging sets of letters\")\n",
    "    top_permutation3, top_score3 = exchange_letters(fixed_letters3, fixed_letter_indices3, letters24, keys24, \n",
    "                                                    Factors24x24, bigrams, bigram_frequencies, verbose=True)\n",
    "\n",
    "    top_layouts.append(top_permutation3)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Optimized layouts (outcome of above)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "load_top_layouts = True\n",
    "print_layouts = False\n",
    "if load_top_layouts:\n",
    "    top_layouts = [\n",
    "    ['Z','H','O','X','P','I','E','A','F','Y','Q','U','M','D','B','J','L','S','N','R','T','C','V','G'],\n",
    "    ['Z','H','O','X','P','I','E','A','F','Y','Q','U','M','L','B','J','D','S','N','R','T','C','V','G'],\n",
    "    ['Z','H','O','X','T','I','E','A','F','Y','Q','U','M','L','P','B','C','S','N','R','D','G','J','V'],\n",
    "    ['Z','H','O','X','P','I','E','A','F','Y','Q','U','M','D','B','J','T','S','N','R','C','L','V','G'],\n",
    "    ['Y','H','O','X','I','P','E','A','Z','F','Q','U','M','D','B','J','L','S','N','R','T','C','V','G'],\n",
    "    ['Y','H','O','X','I','P','E','A','Z','F','Q','U','M','L','B','J','D','S','N','R','T','C','V','G'],\n",
    "    ['Y','H','O','X','I','T','E','A','Z','F','Q','U','M','L','P','B','C','S','N','R','D','G','J','V'],\n",
    "    ['Y','H','O','X','I','P','E','A','Z','F','Q','U','M','D','B','J','T','S','N','R','C','L','V','G'],\n",
    "    ['Y','O','B','F','I','U','E','A','Q','Z','X','G','M','D','P','J','L','S','N','R','T','C','H','V'],\n",
    "    ['Y','O','B','F','I','U','E','A','Q','Z','X','G','M','L','P','J','D','S','N','R','T','C','H','V'],\n",
    "    ['H','O','B','Y','I','U','E','A','Q','Z','X','F','M','D','P','J','C','S','N','R','T','L','V','G'],\n",
    "    ['H','O','B','Y','I','U','E','A','Q','Z','X','F','M','D','P','J','T','S','N','R','C','L','V','G'],\n",
    "    ['Y','Z','H','X','I','O','E','A','P','F','Q','U','M','D','B','J','L','S','N','R','T','C','V','G'],\n",
    "    ['Y','Z','B','X','I','O','E','A','P','F','Q','U','M','L','G','J','D','S','N','R','T','C','H','V'],\n",
    "    ['Y','Z','B','X','I','O','E','A','P','F','Q','U','M','D','H','J','C','S','N','R','T','L','V','G'],\n",
    "    ['Y','Z','B','X','I','O','E','A','P','F','Q','U','M','D','G','J','T','S','N','R','C','L','V','H']]\n",
    "if print_layouts:\n",
    "    print('Layouts:\\n')\n",
    "    for layout in top_layouts:\n",
    "        print(layout)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Rank optimized layouts"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "# %load code/rank_layouts.py\n",
    "layout_strings = []\n",
    "scores = []\n",
    "for layout in top_layouts:\n",
    "    layout_string = ' '.join(layout)\n",
    "    score = score_layout(Factors24x24, layout, bigrams, bigram_frequencies, verbose=False)\n",
    "    #print('    {0}    {1}'.format(layout_string, score))\n",
    "    layout_strings.append(layout_string)\n",
    "    scores.append(score)\n",
    "\n",
    "# Establish which layouts are within a small difference of the top-scoring layout \n",
    "scores_sorted, ranks_sorted, Isort = rank_within_epsilon(scores, factor24, factor=True, verbose=False)\n",
    "layouts_sorted = []\n",
    "layout_strings_sorted = []\n",
    "for i in Isort:\n",
    "    layouts_sorted.append(top_layouts[i])\n",
    "    layout_strings_sorted.append(layout_strings[i])\n",
    "print('\\n    Rank                                                   Score')\n",
    "for i, rank in enumerate(ranks_sorted):\n",
    "    print('    {0}:  {1}    {2}'.format(rank, layout_strings_sorted[i], scores_sorted[i]))\n",
    "\n",
    "print('\\nLayouts tied for first place, with relative letter frequencies:\\n')\n",
    "#print('    Rank                                                   Score')\n",
    "first_ranks = []\n",
    "first_layouts = []\n",
    "first_layout_strings = []\n",
    "first_scores = []\n",
    "for i, rank in enumerate(ranks_sorted):\n",
    "    if rank == 1:\n",
    "        first_ranks.append(rank)\n",
    "        first_layouts.append(layout_strings_sorted[i])\n",
    "        first_layout_strings.append(layouts_sorted[i])\n",
    "        first_scores.append(scores_sorted[i])    \n",
    "Isort2 = np.argsort([-x for x in first_scores])\n",
    "first_ranks_sorted = []\n",
    "first_layouts_sorted = []\n",
    "first_layout_strings_sorted = []\n",
    "first_scores_sorted = []\n",
    "for i in Isort2:\n",
    "    first_ranks_sorted.append(first_ranks[i])\n",
    "    first_layouts_sorted.append(first_layouts[i])\n",
    "    first_layout_strings_sorted.append(first_layout_strings[i])\n",
    "    first_scores_sorted.append(first_scores[i])\n",
    "\n",
    "# Print layouts:\n",
    "for i, layout_string in enumerate(first_layout_strings_sorted):\n",
    "    layout = first_layouts_sorted[i]\n",
    "    print('    Layout {0}:\\n'.format(Isort2[i] + 1))\n",
    "    print_layout24(layout_string)\n",
    "    print('')\n",
    "    print_layout24_instances(layout_string, letters24, instances24, bigrams, bigram_frequencies)\n",
    "    print('')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Ranked, optimized layouts\n",
    "\n",
    "    Rank                                                   Score\n",
    "    1:  Z H O X P I E A F Y Q U M D B J T S N R C L V G    0.718175779085395\n",
    "    2:  Z H O X P I E A F Y Q U M L B J D S N R T C V G    0.7181082154436876\n",
    "    2:  Z H O X P I E A F Y Q U M D B J L S N R T C V G    0.7180926323971318\n",
    "    3:  Y H O X I P E A Z F Q U M D B J T S N R C L V G    0.7180491746992222\n",
    "    4:  Z H O X T I E A F Y Q U M L P B C S N R D G J V    0.7179953118351015\n",
    "    4:  Y H O X I P E A Z F Q U M L B J D S N R T C V G    0.7179816110575149\n",
    "    4:  Y H O X I P E A Z F Q U M D B J L S N R T C V G    0.717966028010959\n",
    "    5:  Y H O X I T E A Z F Q U M L P B C S N R D G J V    0.7178848621521811\n",
    "    6:  Y Z B X I O E A P F Q U M D G J T S N R C L V H    0.7176201586204008\n",
    "    6:  H O B Y I U E A Q Z X F M D P J T S N R C L V G    0.7176038179852999\n",
    "    6:  Y Z B X I O E A P F Q U M D H J C S N R T L V G    0.7175719371257249\n",
    "    7:  H O B Y I U E A Q Z X F M D P J C S N R T L V G    0.7175637424926007\n",
    "    7:  Y Z B X I O E A P F Q U M L G J D S N R T C H V    0.717538426789943\n",
    "    7:  Y Z H X I O E A P F Q U M D B J L S N R T C V G    0.7175223166015284\n",
    "    7:  Y O B F I U E A Q Z X G M L P J D S N R T C H V    0.7175175913298815\n",
    "    8:  Y O B F I U E A Q Z X G M D P J L S N R T C H V    0.7174898077178882\n",
    "\n",
    "Unique letters per column:\n",
    "\n",
    "        Z H O X P I E A F Y Q U M D B J T S N R C L V G\n",
    "        Y Z B Y I P     Z F X F   L P B D       T C J V\n",
    "        H O   F T T     P Z   G     G   L       D G H H\n",
    "                  O     Q           P   C\n",
    "                  U                 H\n",
    "\n",
    "        - - - - - - E A - - - - M - - - - S N R - - - -    >1 letter per column\n",
    "\n",
    "Unique letters per column for the top 8 rows (5 places):\n",
    "\n",
    "        Z H O X P I E A F Y Q U M D B J T S N R C L V G\n",
    "        Y       I P     Z F       L P B D       T C J V\n",
    "                T T                     L       D G    \n",
    "                                        C\n",
    "                                     \n",
    "Unique letters per column for the top 4 rows (3 places):\n",
    "\n",
    "        Z H O X P I E A F Y Q U M D B J T S N R C L V G\n",
    "        Y       I P     Z F       L     D       T C  \n",
    "                                        L              \n",
    "                                         \n",
    "        - H O X - - E A - - Q U M - B J - S N R - - V G    >1 letter per column in top 4 rows\n",
    "          \n",
    "\n",
    "Layouts tied for first place, with relative letter frequencies:\n",
    "\n",
    "    Layout 1:\n",
    "\n",
    "    Z H O X  M D B J\n",
    "    P I E A  T S N R\n",
    "    F Y Q U  C L V G\n",
    "\n",
    "     49 101 1142  22  350 638 160  59\n",
    "    344 837 1695 1553  565 854 918 859\n",
    "    103 104 112 525  559 714 126 150\n",
    "\n",
    "    left: 658688805 (10.67%)  right: 595172561\n",
    "    Total same-finger bigram frequencies:     48977687406\n",
    "    Total bigram inward roll frequencies:   5838991231649\n",
    "    \n",
    "\n",
    "### Create variants of top-scoring layouts\n",
    "\n",
    "As an alternative to simply choosing the first-place layout, we can generate variations of this layout and find those variants within a small difference of one another and select from among these variants. For this, we select keys to vary, compute scores for every combination of the letters assigned to these keys, and select among those that are tied for first place. Since no other layout is tied for first place, we will vary those keys with different letters in the top four layouts (in first, second, and third place):\n",
    "\n",
    "        - H O X  M - B J\n",
    "        - - E A  - S N R \n",
    "        - - Q U  - - V G\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "#   - H O X  M - B J\n",
    "#   - - E A  - S N R \n",
    "#   - - Q U  - - V G\n",
    "\n",
    "fixed_letters = ['H','O','X', 'E','A', 'Q','U', 'M','B','J', 'S','N','R', 'V','G']\n",
    "fixed_letter_indices = [1,2,3, 6,7, 10,11, 12,14,15, 17,18,19, 22,23]\n",
    "open_letter_indices  = [0, 4,5, 8,9, 13, 16, 20,21]\n",
    "\n",
    "top_variant_permutation, top_variant_score, variant_letter_permutations = permute_optimize_keys(fixed_letters, \n",
    "    fixed_letter_indices, open_letter_indices, letters24, keys24, Factors24x24, \n",
    "    bigrams, bigram_frequencies, min_score=0, verbose=False)\n",
    "\n",
    "variant_scores = []\n",
    "nletters = len(fixed_letter_indices) + len(open_letter_indices)\n",
    "layout_variant_strings = []\n",
    "for ipermutation, letter_permutation in enumerate(variant_letter_permutations):\n",
    "    variant_letters = np.array(['W' for x in range(nletters)])  # KEEP to initialize!\n",
    "    for imove, open_letter_index in enumerate(open_letter_indices):\n",
    "        variant_letters[open_letter_index] = letter_permutation[imove]\n",
    "    for ifixed, fixed_letter_index in enumerate(fixed_letter_indices):\n",
    "        variant_letters[fixed_letter_index] = fixed_letters[ifixed]\n",
    "    layout_variant_strings.append(variant_letters)\n",
    "    # Compute the score for this permutation:\n",
    "    variant_score = score_layout(Factors24x24, variant_letters, \n",
    "                                 bigrams, bigram_frequencies, verbose=False)\n",
    "    variant_scores.append(variant_score)\n",
    "\n",
    "layout_variants = []\n",
    "for layout_string in layout_variant_strings:\n",
    "    layout = ' '.join(layout_string)\n",
    "    layout_variants.append(layout)\n",
    "\n",
    "variant_scores_sorted, variant_ranks_sorted, Isort_variants = rank_within_epsilon(variant_scores, \n",
    "        factor24, factor=True, verbose=False)\n",
    "layout_variants_sorted = []\n",
    "layout_variant_strings_sorted = []\n",
    "for i in Isort_variants:\n",
    "    layout_variants_sorted.append(layout_variants[i])\n",
    "    layout_variant_strings_sorted.append(layout_variant_strings[i])\n",
    "\n",
    "print('    Rank                                                   Score')\n",
    "for i, rank in enumerate(variant_ranks_sorted):\n",
    "    if rank == 1:\n",
    "        print('    {0}:  {1}    {2}'.format(rank, layout_variants_sorted[i], variant_scores_sorted[i]))\n",
    "\n",
    "# Print layouts:\n",
    "Ifirst_place = []\n",
    "layout_variants_first_place = []\n",
    "layout_variant_strings_first_place = []\n",
    "for i, rank in enumerate(variant_ranks_sorted):\n",
    "    if rank == 1:\n",
    "        layout_string = layout_variant_strings_sorted[i]\n",
    "        layout = layout_variants_sorted[i]\n",
    "        print('\\n    Layout {0}:\\n'.format(i + 1))\n",
    "        print_layout24(layout_string)\n",
    "        print('')\n",
    "        print_layout24_instances(layout_string, letters24, instances24, \n",
    "                                 bigrams, bigram_frequencies)\n",
    "\n",
    "        Ifirst_place.append(i)\n",
    "        layout_variants_first_place.append(layout)\n",
    "        layout_variant_strings_first_place.append(layout_string)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "####   Variants of top-scoring layouts tied for first place\n",
    "\n",
    "    Rank                                                   Score\n",
    "    1:  Z H O X P I E A F Y Q U M D B J T S N R C L V G    0.718175779085395\n",
    "    1:  Y H O X P I E A F Z Q U M D B J T S N R C L V G    0.7181740824015593\n",
    "    1:  Z H O X P I E A F Y Q U M C B J T S N R D L V G    0.7181662126172352\n",
    "    1:  Y H O X P I E A F Z Q U M C B J T S N R D L V G    0.7181645159333996\n",
    "    1:  Z H O X P I E A F Y Q U M L B J T S N R C D V G    0.7181440154199303\n",
    "    1:  Y H O X P I E A F Z Q U M L B J T S N R C D V G    0.7181423187360948\n",
    "    1:  Z H O X P I E A F Y Q U M L B J T S N R D C V G    0.7181399906496914\n",
    "    1:  Y H O X P I E A F Z Q U M L B J T S N R D C V G    0.7181382939658557\n",
    "    1:  Z H O X P I E A F Y Q U M D B J C S N R T L V G    0.7181368885518766\n",
    "    1:  Y H O X P I E A F Z Q U M D B J C S N R T L V G    0.7181351918680411\n",
    "    1:  Y H O X P I E A Z F Q U M D B J T S N R C L V G    0.7181315366108767\n",
    "    1:  F H O X P I E A Z Y Q U M D B J T S N R C L V G    0.7181279794445848\n",
    "\n",
    "    Layout 1:\n",
    "\n",
    "    Z H O X  M D B J\n",
    "    P I E A  T S N R\n",
    "    F Y Q U  C L V G\n",
    "\n",
    "     49 101 1142  22  350 638 160  59\n",
    "    344 837 1695 1553  565 854 918 859\n",
    "    103 104 112 525  559 714 126 150\n",
    "\n",
    "    left: 658688805 (10.67%)  right: 595172561\n",
    "    Total same-finger bigram frequencies:     48977687406\n",
    "    Total bigram inward roll frequencies:   5838991231649\n",
    "\n",
    "    Layout 2:\n",
    "\n",
    "    Y H O X  M D B J\n",
    "    P I E A  T S N R\n",
    "    F Z Q U  C L V G\n",
    "\n",
    "    104 101 1142  22  350 638 160  59\n",
    "    344 837 1695 1553  565 854 918 859\n",
    "    103  49 112 525  559 714 126 150\n",
    "\n",
    "    left: 658688805 (10.67%)  right: 595172561\n",
    "    Total same-finger bigram frequencies:     52923154294\n",
    "    Total bigram inward roll frequencies:   5835045764762\n",
    "\n",
    "    Layout 3:\n",
    "\n",
    "    Z H O X  M C B J\n",
    "    P I E A  T S N R\n",
    "    F Y Q U  D L V G\n",
    "\n",
    "     49 101 1142  22  350 559 160  59\n",
    "    344 837 1695 1553  565 854 918 859\n",
    "    103 104 112 525  638 714 126 150\n",
    "\n",
    "    left: 658688805 (10.67%)  right: 595172561\n",
    "    Total same-finger bigram frequencies:     50643072440\n",
    "    Total bigram inward roll frequencies:   5837325846616\n",
    "\n",
    "    Layout 4:\n",
    "\n",
    "    Y H O X  M C B J\n",
    "    P I E A  T S N R\n",
    "    F Z Q U  D L V G\n",
    "\n",
    "    104 101 1142  22  350 559 160  59\n",
    "    344 837 1695 1553  565 854 918 859\n",
    "    103  49 112 525  638 714 126 150\n",
    "\n",
    "    left: 658688805 (10.67%)  right: 595172561\n",
    "    Total same-finger bigram frequencies:     54588539327\n",
    "    Total bigram inward roll frequencies:   5833380379728\n",
    "\n",
    "    Layout 5:\n",
    "\n",
    "    Z H O X  M L B J\n",
    "    P I E A  T S N R\n",
    "    F Y Q U  C D V G\n",
    "\n",
    "     49 101 1142  22  350 714 160  59\n",
    "    344 837 1695 1553  565 854 918 859\n",
    "    103 104 112 525  559 638 126 150\n",
    "\n",
    "    left: 658688805 (10.67%)  right: 595172561\n",
    "    Total same-finger bigram frequencies:     48977687406\n",
    "    Total bigram inward roll frequencies:   5838991231649\n",
    "\n",
    "    Layout 6:\n",
    "\n",
    "    Y H O X  M L B J\n",
    "    P I E A  T S N R\n",
    "    F Z Q U  C D V G\n",
    "\n",
    "    104 101 1142  22  350 714 160  59\n",
    "    344 837 1695 1553  565 854 918 859\n",
    "    103  49 112 525  559 638 126 150\n",
    "\n",
    "    left: 658688805 (10.67%)  right: 595172561\n",
    "    Total same-finger bigram frequencies:     52923154294\n",
    "    Total bigram inward roll frequencies:   5835045764762\n",
    "\n",
    "    Layout 7:\n",
    "\n",
    "    Z H O X  M L B J\n",
    "    P I E A  T S N R\n",
    "    F Y Q U  D C V G\n",
    "\n",
    "     49 101 1142  22  350 714 160  59\n",
    "    344 837 1695 1553  565 854 918 859\n",
    "    103 104 112 525  638 559 126 150\n",
    "\n",
    "    left: 658688805 (10.67%)  right: 595172561\n",
    "    Total same-finger bigram frequencies:     50643072440\n",
    "    Total bigram inward roll frequencies:   5837325846616\n",
    "\n",
    "    Layout 8:\n",
    "\n",
    "    Y H O X  M L B J\n",
    "    P I E A  T S N R\n",
    "    F Z Q U  D C V G\n",
    "\n",
    "    104 101 1142  22  350 714 160  59\n",
    "    344 837 1695 1553  565 854 918 859\n",
    "    103  49 112 525  638 559 126 150\n",
    "\n",
    "    left: 658688805 (10.67%)  right: 595172561\n",
    "    Total same-finger bigram frequencies:     54588539327\n",
    "    Total bigram inward roll frequencies:   5833380379728\n",
    "\n",
    "    Layout 9:\n",
    "\n",
    "    Z H O X  M D B J\n",
    "    P I E A  C S N R\n",
    "    F Y Q U  T L V G\n",
    "\n",
    "     49 101 1142  22  350 638 160  59\n",
    "    344 837 1695 1553  559 854 918 859\n",
    "    103 104 112 525  565 714 126 150\n",
    "\n",
    "    left: 658688805 (10.67%)  right: 595172561\n",
    "    Total same-finger bigram frequencies:     48977687406\n",
    "    Total bigram inward roll frequencies:   5838991231649\n",
    "\n",
    "    Layout 10:\n",
    "\n",
    "    Y H O X  M D B J\n",
    "    P I E A  C S N R\n",
    "    F Z Q U  T L V G\n",
    "\n",
    "    104 101 1142  22  350 638 160  59\n",
    "    344 837 1695 1553  559 854 918 859\n",
    "    103  49 112 525  565 714 126 150\n",
    "\n",
    "    left: 658688805 (10.67%)  right: 595172561\n",
    "    Total same-finger bigram frequencies:     52923154294\n",
    "    Total bigram inward roll frequencies:   5835045764762\n",
    "\n",
    "    Layout 11:\n",
    "\n",
    "    Y H O X  M D B J\n",
    "    P I E A  T S N R\n",
    "    Z F Q U  C L V G\n",
    "\n",
    "    104 101 1142  22  350 638 160  59\n",
    "    344 837 1695 1553  565 854 918 859\n",
    "     49 103 112 525  559 714 126 150\n",
    "\n",
    "    left: 658688805 (10.67%)  right: 595172561\n",
    "    Total same-finger bigram frequencies:     59523779284\n",
    "    Total bigram inward roll frequencies:   5828445139772\n",
    "\n",
    "    Layout 12:\n",
    "\n",
    "    F H O X  M D B J\n",
    "    P I E A  T S N R\n",
    "    Z Y Q U  C L V G\n",
    "\n",
    "    103 101 1142  22  350 638 160  59\n",
    "    344 837 1695 1553  565 854 918 859\n",
    "     49 104 112 525  559 714 126 150\n",
    "\n",
    "    left: 658688805 (10.67%)  right: 595172561\n",
    "    Total same-finger bigram frequencies:     48977687406\n",
    "    Total bigram inward roll frequencies:   5838991231649"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "layout_variant_strings_first_place = [\n",
    "    ['Z','H','O','X','P','I','E','A','F','Y','Q','U','M','D','B','J','T','S','N','R','C','L','V','G'],\n",
    "    ['Y','H','O','X','P','I','E','A','F','Z','Q','U','M','D','B','J','T','S','N','R','C','L','V','G'],\n",
    "    ['Z','H','O','X','P','I','E','A','F','Y','Q','U','M','C','B','J','T','S','N','R','D','L','V','G'],\n",
    "    ['Y','H','O','X','P','I','E','A','F','Z','Q','U','M','C','B','J','T','S','N','R','D','L','V','G'],\n",
    "    ['Z','H','O','X','P','I','E','A','F','Y','Q','U','M','L','B','J','T','S','N','R','C','D','V','G'],\n",
    "    ['Y','H','O','X','P','I','E','A','F','Z','Q','U','M','L','B','J','T','S','N','R','C','D','V','G'],\n",
    "    ['Z','H','O','X','P','I','E','A','F','Y','Q','U','M','L','B','J','T','S','N','R','D','C','V','G'],\n",
    "    ['Y','H','O','X','P','I','E','A','F','Z','Q','U','M','L','B','J','T','S','N','R','D','C','V','G'],\n",
    "    ['Z','H','O','X','P','I','E','A','F','Y','Q','U','M','D','B','J','C','S','N','R','T','L','V','G'],\n",
    "    ['Y','H','O','X','P','I','E','A','F','Z','Q','U','M','D','B','J','C','S','N','R','T','L','V','G'],\n",
    "    ['Y','H','O','X','P','I','E','A','Z','F','Q','U','M','D','B','J','T','S','N','R','C','L','V','G'],\n",
    "    ['F','H','O','X','P','I','E','A','Z','Y','Q','U','M','D','B','J','T','S','N','R','C','L','V','G']]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Step 4: Evaluate winning layout <a name=\"step4\">\n",
    "    \n",
    "We evaluate the candidate winner with tests:\n",
    "    \n",
    "    1. Evaluate variants of the candidate winner using interkey speed estimates\n",
    "    2. Evaluate sensitivity of the variants to the scoring parameters\n",
    "    3. Search for higher-scoring layouts by rearranging letters"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Test 1. Evaluate variants of the candidate winner using interkey speed estimates  \n",
    "Below we rescore the four winning variants that are tied for first place, replacing the factor matrix with the inter-key speed matrix.\n",
    "\n",
    "*Note:*\n",
    "    \n",
    "The speed matrix contains normalized interkey stroke times derived from a published study (\"Estimation of digraph costs for keyboard layout optimization\", A Iseri, Ma Eksioglu, International Journal of Industrial Ergonomics, 48, 127-138, 2015). To establish which layouts are within a small difference of each other when using the speed matrix, we define an epsilon equal to 131.58 ms for a single bigram (of the 32^2 possible bigrams), where 131.58 ms is the fastest measured digraph tapping speed (30,000/228 = 131.58 ms) recorded in the above study.\n",
    "    \n",
    "\"Digraph-tapping rate changes dramatically across the digraph types. The range is between 82 and 228 taps per 30 s. The difference is nearly three times between the slowest and the fastest digraphs. From this result it can be concluded that the assignment of letter pairs on the correct digraph keys on the keyboard can have a high impact on the typing speed.\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_layout_strings = layout_variant_strings_first_place"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# %load code/test/score_speed_of_layouts.py\n",
    "data_matrix_speed = Speed24x24  # SpeedSymmetric24x24\n",
    "speed_scores = []\n",
    "for letters in test_layout_strings:\n",
    "    score = score_layout(data_matrix_speed, letters,  bigrams, bigram_frequencies, verbose = False) \n",
    "    speed_scores.append(score)\n",
    "\n",
    "speed_scores_sorted, speed_ranks_sorted, Isort_speed = rank_within_epsilon(speed_scores, \n",
    "                                                                    epsilon, factor=False, verbose=False)\n",
    "speed_layouts_sorted = []\n",
    "speed_layout_strings_sorted = []\n",
    "for i in Isort_speed:\n",
    "    speed_layouts_sorted.append(' '.join(test_layout_strings[i]))\n",
    "    speed_layout_strings_sorted.append(test_layout_strings[i])\n",
    "\n",
    "count = 0\n",
    "print('    Layout                                                  Speed score')\n",
    "for i, isort_speed in enumerate(Isort_speed):\n",
    "    if speed_ranks_sorted[isort_speed] == 1:\n",
    "        count += 1\n",
    "        if isort_speed < 9:\n",
    "            s = '  '\n",
    "        else:\n",
    "            s = ' '\n",
    "        print('    ({0}){1}{2}    {3}'.format(isort_speed+1, s, \n",
    "                                              speed_layouts_sorted[i], \n",
    "                                              speed_scores_sorted[i]))\n",
    "print('    {0} of {1} layouts tied for first place'.format(count, len(test_layout_strings)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Variants 1-10 tie for top score:\n",
    "\n",
    "    Layout                                                  Speed score\n",
    "    (8)  Y H O X P I E A F Z Q U M L B J T S N R D C V G    0.7126582048217771\n",
    "    (7)  Z H O X P I E A F Y Q U M L B J T S N R D C V G    0.7126537342949005\n",
    "    (6)  Y H O X P I E A F Z Q U M L B J T S N R C D V G    0.7126482055890047\n",
    "    (4)  Y H O X P I E A F Z Q U M C B J T S N R D L V G    0.7126473432931382\n",
    "    (5)  Z H O X P I E A F Y Q U M L B J T S N R C D V G    0.7126437138130693\n",
    "    (3)  Z H O X P I E A F Y Q U M C B J T S N R D L V G    0.7126428727662616\n",
    "    (2)  Y H O X P I E A F Z Q U M D B J T S N R C L V G    0.7126324844095655\n",
    "    (1)  Z H O X P I E A F Y Q U M D B J T S N R C L V G    0.7126279926336301\n",
    "    (10) Y H O X P I E A F Z Q U M D B J C S N R T L V G    0.7126258758990177\n",
    "    (9)  Z H O X P I E A F Y Q U M D B J C S N R T L V G    0.7126213841230822\n",
    "    10 of 12 layouts tied for first place"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Test 2. Evaluate sensitivity of the candidate winner to the scoring parameters\n",
    "\n",
    "We run a test below on the variants to see how robust they are to removal of scoring parameters. We removed each of the 11 scoring parameters one by one and ranked the new scores for the variants above. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# %load code/test/remove_parameters_rescore.py\n",
    "params0 = [side_above_3away, side_above_2away, side_above_1away, middle_above_ring, ring_above_middle, \n",
    "           outward, skip_row_3away, skip_row_2away, skip_row_1away, skip_row_0away, same_finger]\n",
    "param_names = ['side_above_3away', 'side_above_2away', 'side_above_1away', \n",
    "               'middle_above_ring', 'ring_above_middle', 'outward', 'skip_row_3away', \n",
    "               'skip_row_2away', 'skip_row_1away', 'skip_row_0away', 'same_finger']\n",
    "params_lists = []\n",
    "for i in range(len(params0)):\n",
    "    params_list = params0.copy()\n",
    "    params_list[i] = 1.0\n",
    "    params_lists.append(params_list)\n",
    "\n",
    "for iparam, P in enumerate(params_lists):\n",
    "\n",
    "    print('    Remove parameter {0}:'.format(param_names[iparam]))\n",
    "\n",
    "    data_matrix_param = create_24x24_flow_matrix(not_home_row, side_top,\n",
    "                                                 P[0],P[1],P[2],P[3],P[4],P[5],P[6],P[7],P[8],P[9],P[10],\n",
    "                                                 1,1,1,1,1,1)\n",
    "    if apply_strength:\n",
    "        data_matrix_param = Strength24x24 * data_matrix_param\n",
    "\n",
    "    param_scores = []\n",
    "    for letters in test_layout_strings:\n",
    "        score = score_layout(data_matrix_param, letters, bigrams, bigram_frequencies, verbose=False);\n",
    "        param_scores.append(score)\n",
    "            \n",
    "    param_scores_sorted, param_ranks_sorted, Isort_param = rank_within_epsilon(param_scores, factor24, factor=True, verbose=False)\n",
    "    param_layouts_sorted = []\n",
    "    param_layout_strings_sorted = []\n",
    "    for i in Isort_param:\n",
    "        param_layouts_sorted.append(' '.join(test_layout_strings[i]))\n",
    "        param_layout_strings_sorted.append(test_layout_strings[i])\n",
    "\n",
    "    print('    Variant                                                 Score')\n",
    "    count = 0\n",
    "    for i, isort_param in enumerate(Isort_param):\n",
    "        count += 1\n",
    "        if param_ranks_sorted[isort_param] == 1:\n",
    "            if isort_param < 9:\n",
    "                s = '  '\n",
    "            else:\n",
    "                s = ' '\n",
    "            print('    ({0}){1}{2}    {3}'.format(isort_param+1, s, \n",
    "                                                  param_layouts_sorted[i], \n",
    "                                                  param_scores_sorted[i]))\n",
    "    print('    {0} of {1} layouts tied for first place'.format(count, len(test_layout_strings)))"
   ]
  },
  {
   "cell_type": "raw",
   "metadata": {},
   "source": [
    "Variant 1 scores higher than any other most often, demonstrating that this layout is among the least sensitive to individual parameters.\n",
    "\n",
    "    Remove parameter side_above_3away:\n",
    "    Variant                                                 Score\n",
    "    (9)  Z H O X P I E A F Y Q U M D B J C S N R T L V G    0.7210583773221152\n",
    "    (1)  Z H O X P I E A F Y Q U M D B J T S N R C L V G    0.721056939210811\n",
    "    (10) Y H O X P I E A F Z Q U M D B J C S N R T L V G    0.7210527642648356\n",
    "    (2)  Y H O X P I E A F Z Q U M D B J T S N R C L V G    0.7210513261535314\n",
    "    (3)  Z H O X P I E A F Y Q U M C B J T S N R D L V G    0.7210467994285263\n",
    "    (4)  Y H O X P I E A F Z Q U M C B J T S N R D L V G    0.7210411863712467\n",
    "    (5)  Z H O X P I E A F Y Q U M L B J T S N R C D V G    0.7210251755453465\n",
    "    (7)  Z H O X P I E A F Y Q U M L B J T S N R D C V G    0.7210205774609825\n",
    "    (6)  Y H O X P I E A F Z Q U M L B J T S N R C D V G    0.7210195624880669\n",
    "    (11) Y H O X P I E A Z F Q U M D B J T S N R C L V G    0.7210172538163603\n",
    "    (12) F H O X P I E A Z Y Q U M D B J T S N R C L V G    0.721016520203857\n",
    "    (8)  Y H O X P I E A F Z Q U M L B J T S N R D C V G    0.7210149644037028\n",
    "    12 of 12 layouts tied for first place\n",
    "    Remove parameter side_above_2away:\n",
    "    Variant                                                 Score\n",
    "    (2)  Y H O X P I E A F Z Q U M D B J T S N R C L V G    0.7233214925197848\n",
    "    (1)  Z H O X P I E A F Y Q U M D B J T S N R C L V G    0.7233163652876\n",
    "    (4)  Y H O X P I E A F Z Q U M C B J T S N R D L V G    0.7233119260516252\n",
    "    (6)  Y H O X P I E A F Z Q U M L B J T S N R C D V G    0.7233094386419149\n",
    "    (3)  Z H O X P I E A F Y Q U M C B J T S N R D L V G    0.7233067988194403\n",
    "    (8)  Y H O X P I E A F Z Q U M L B J T S N R D C V G    0.7233065402026811\n",
    "    (5)  Z H O X P I E A F Y Q U M L B J T S N R C D V G    0.7233043114097301\n",
    "    (7)  Z H O X P I E A F Y Q U M L B J T S N R D C V G    0.7233014129704961\n",
    "    (10) Y H O X P I E A F Z Q U M D B J C S N R T L V G    0.7232825203140769\n",
    "    (9)  Z H O X P I E A F Y Q U M D B J C S N R T L V G    0.7232773930818921\n",
    "    (11) Y H O X P I E A Z F Q U M D B J T S N R C L V G    0.7232625473933775\n",
    "    12 of 12 layouts tied for first place\n",
    "    Remove parameter side_above_1away:\n",
    "    Variant                                                 Score\n",
    "    (1)  Z H O X P I E A F Y Q U M D B J T S N R C L V G    0.7251645905252198\n",
    "    (3)  Z H O X P I E A F Y Q U M C B J T S N R D L V G    0.72515502405706\n",
    "    (2)  Y H O X P I E A F Z Q U M D B J T S N R C L V G    0.7251475103014832\n",
    "    (4)  Y H O X P I E A F Z Q U M C B J T S N R D L V G    0.7251379438333233\n",
    "    (7)  Z H O X P I E A F Y Q U M L B J T S N R D C V G    0.7251345319311451\n",
    "    (8)  Y H O X P I E A F Z Q U M L B J T S N R D C V G    0.7251174517074084\n",
    "    (5)  Z H O X P I E A F Y Q U M L B J T S N R C D V G    0.7251078181722477\n",
    "    (6)  Y H O X P I E A F Z Q U M L B J T S N R C D V G    0.7250907379485111\n",
    "    12 of 12 layouts tied for first place\n",
    "    Remove parameter middle_above_ring:\n",
    "    Variant                                                 Score\n",
    "    (2)  Y H O X P I E A F Z Q U M D B J T S N R C L V G    0.719576614490204\n",
    "    (1)  Z H O X P I E A F Y Q U M D B J T S N R C L V G    0.7195765506408036\n",
    "    (4)  Y H O X P I E A F Z Q U M C B J T S N R D L V G    0.7195556622780318\n",
    "    (3)  Z H O X P I E A F Y Q U M C B J T S N R D L V G    0.7195555984286314\n",
    "    (6)  Y H O X P I E A F Z Q U M L B J T S N R C D V G    0.7195084229854201\n",
    "    (5)  Z H O X P I E A F Y Q U M L B J T S N R C D V G    0.7195083591360196\n",
    "    (8)  Y H O X P I E A F Z Q U M L B J T S N R D C V G    0.719504398215181\n",
    "    (7)  Z H O X P I E A F Y Q U M L B J T S N R D C V G    0.7195043343657805\n",
    "    12 of 12 layouts tied for first place\n",
    "    Remove parameter ring_above_middle:\n",
    "    Variant                                                 Score\n",
    "    (5)  Z H O X P I E A F Y Q U M L B J T S N R C D V G    0.7217739948393724\n",
    "    (6)  Y H O X P I E A F Z Q U M L B J T S N R C D V G    0.7217722981555368\n",
    "    (1)  Z H O X P I E A F Y Q U M D B J T S N R C L V G    0.7217234092956235\n",
    "    (2)  Y H O X P I E A F Z Q U M D B J T S N R C L V G    0.721721712611788\n",
    "    (3)  Z H O X P I E A F Y Q U M C B J T S N R D L V G    0.7217138428274638\n",
    "    (4)  Y H O X P I E A F Z Q U M C B J T S N R D L V G    0.7217121461436282\n",
    "    12 of 12 layouts tied for first place\n",
    "    Remove parameter outward:\n",
    "    Variant                                                 Score\n",
    "    (1)  Z H O X P I E A F Y Q U M D B J T S N R C L V G    0.7304521334502336\n",
    "    (2)  Y H O X P I E A F Z Q U M D B J T S N R C L V G    0.7304470031869887\n",
    "    (3)  Z H O X P I E A F Y Q U M C B J T S N R D L V G    0.7304377416670281\n",
    "    (4)  Y H O X P I E A F Z Q U M C B J T S N R D L V G    0.7304326114037832\n",
    "    (5)  Z H O X P I E A F Y Q U M L B J T S N R C D V G    0.7304150062665732\n",
    "    (6)  Y H O X P I E A F Z Q U M L B J T S N R C D V G    0.7304098760033282\n",
    "    (9)  Z H O X P I E A F Y Q U M D B J C S N R T L V G    0.7304086837530025\n",
    "    (7)  Z H O X P I E A F Y Q U M L B J T S N R D C V G    0.730407663362785\n",
    "    (11) Y H O X P I E A Z F Q U M D B J T S N R C L V G    0.7304062233896591\n",
    "    (12) F H O X P I E A Z Y Q U M D B J T S N R C L V G    0.7304060351998625\n",
    "    (10) Y H O X P I E A F Z Q U M D B J C S N R T L V G    0.7304035534897575\n",
    "    (8)  Y H O X P I E A F Z Q U M L B J T S N R D C V G    0.73040253309954\n",
    "    12 of 12 layouts tied for first place\n",
    "    Remove parameter skip_row_3away:\n",
    "    Variant                                                 Score\n",
    "    (2)  Y H O X P I E A F Z Q U M D B J T S N R C L V G    0.7190492106548195\n",
    "    (1)  Z H O X P I E A F Y Q U M D B J T S N R C L V G    0.7190484323510822\n",
    "    (4)  Y H O X P I E A F Z Q U M C B J T S N R D L V G    0.7190397151951259\n",
    "    (3)  Z H O X P I E A F Y Q U M C B J T S N R D L V G    0.7190389368913884\n",
    "    (6)  Y H O X P I E A F Z Q U M L B J T S N R C D V G    0.719017446989355\n",
    "    (5)  Z H O X P I E A F Y Q U M L B J T S N R C D V G    0.7190166686856175\n",
    "    (8)  Y H O X P I E A F Z Q U M L B J T S N R D C V G    0.719013493227582\n",
    "    (7)  Z H O X P I E A F Y Q U M L B J T S N R D C V G    0.7190127149238446\n",
    "    (10) Y H O X P I E A F Z Q U M D B J C S N R T L V G    0.7190103176768552\n",
    "    (9)  Z H O X P I E A F Y Q U M D B J C S N R T L V G    0.7190095393731178\n",
    "    (12) F H O X P I E A Z Y Q U M D B J T S N R C L V G    0.7190080133441281\n",
    "    (11) Y H O X P I E A Z F Q U M D B J T S N R C L V G    0.7190066554444265\n",
    "    12 of 12 layouts tied for first place\n",
    "    Remove parameter skip_row_2away:\n",
    "    Variant                                                 Score\n",
    "    (1)  Z H O X P I E A F Y Q U M D B J T S N R C L V G    0.723026120277833\n",
    "    (2)  Y H O X P I E A F Z Q U M D B J T S N R C L V G    0.7230237186383444\n",
    "    (3)  Z H O X P I E A F Y Q U M C B J T S N R D L V G    0.7230162336491563\n",
    "    (4)  Y H O X P I E A F Z Q U M C B J T S N R D L V G    0.7230138320096677\n",
    "    (5)  Z H O X P I E A F Y Q U M L B J T S N R C D V G    0.7230098079422689\n",
    "    (6)  Y H O X P I E A F Z Q U M L B J T S N R C D V G    0.7230074063027803\n",
    "    (7)  Z H O X P I E A F Y Q U M L B J T S N R D C V G    0.7230056187074542\n",
    "    (8)  Y H O X P I E A F Z Q U M L B J T S N R D C V G    0.7230032170679657\n",
    "    (9)  Z H O X P I E A F Y Q U M D B J C S N R T L V G    0.7229884543275135\n",
    "    (10) Y H O X P I E A F Z Q U M D B J C S N R T L V G    0.7229860526880248\n",
    "    (11) Y H O X P I E A Z F Q U M D B J T S N R C L V G    0.7229755076488105\n",
    "    12 of 12 layouts tied for first place\n",
    "    Remove parameter skip_row_1away:\n",
    "    Variant                                                 Score\n",
    "    (1)  Z H O X P I E A F Y Q U M D B J T S N R C L V G    0.7285345017489695\n",
    "    (2)  Y H O X P I E A F Z Q U M D B J T S N R C L V G    0.7285267361659309\n",
    "    (3)  Z H O X P I E A F Y Q U M C B J T S N R D L V G    0.7285234335560208\n",
    "    (4)  Y H O X P I E A F Z Q U M C B J T S N R D L V G    0.7285156679729821\n",
    "    (9)  Z H O X P I E A F Y Q U M D B J C S N R T L V G    0.7284954366921329\n",
    "    (5)  Z H O X P I E A F Y Q U M L B J T S N R C D V G    0.7284940709462921\n",
    "    (10) Y H O X P I E A F Z Q U M D B J C S N R T L V G    0.7284876711090943\n",
    "    (6)  Y H O X P I E A F Z Q U M L B J T S N R C D V G    0.7284863053632535\n",
    "    (7)  Z H O X P I E A F Y Q U M L B J T S N R D C V G    0.7284704290580903\n",
    "    (8)  Y H O X P I E A F Z Q U M L B J T S N R D C V G    0.7284626634750516\n",
    "    12 of 12 layouts tied for first place\n",
    "    Remove parameter skip_row_0away:\n",
    "    Variant                                                 Score\n",
    "    (3)  Z H O X P I E A F Y Q U M C B J T S N R D L V G    0.7223799823972282\n",
    "    (1)  Z H O X P I E A F Y Q U M D B J T S N R C L V G    0.7223782546378588\n",
    "    (4)  Y H O X P I E A F Z Q U M C B J T S N R D L V G    0.7223782461644987\n",
    "    (2)  Y H O X P I E A F Z Q U M D B J T S N R C L V G    0.7223765184051292\n",
    "    (7)  Z H O X P I E A F Y Q U M L B J T S N R D C V G    0.7223537604296844\n",
    "    (8)  Y H O X P I E A F Z Q U M L B J T S N R D C V G    0.7223520241969549\n",
    "    (5)  Z H O X P I E A F Y Q U M L B J T S N R C D V G    0.7223464909723941\n",
    "    (6)  Y H O X P I E A F Z Q U M L B J T S N R C D V G    0.7223447547396646\n",
    "    (9)  Z H O X P I E A F Y Q U M D B J C S N R T L V G    0.7223396437775654\n",
    "    (10) Y H O X P I E A F Z Q U M D B J C S N R T L V G    0.722337907544836\n",
    "    (11) Y H O X P I E A Z F Q U M D B J T S N R C L V G    0.7223339704955882\n",
    "    (12) F H O X P I E A Z Y Q U M D B J T S N R C L V G    0.7223304549970486\n",
    "    12 of 12 layouts tied for first place\n",
    "    Remove parameter same_finger:\n",
    "    Variant                                                 Score\n",
    "    (2)  Y H O X P I E A F Z Q U M D B J T S N R C L V G    0.7409975074251155\n",
    "    (4)  Y H O X P I E A F Z Q U M C B J T S N R D L V G    0.7409927982917923\n",
    "    (1)  Z H O X P I E A F Y Q U M D B J T S N R C L V G    0.7409730432979155\n",
    "    (3)  Z H O X P I E A F Y Q U M C B J T S N R D L V G    0.7409683341645922\n",
    "    (8)  Y H O X P I E A F Z Q U M L B J T S N R D C V G    0.7409665763242484\n",
    "    (6)  Y H O X P I E A F Z Q U M L B J T S N R C D V G    0.7409657437596509\n",
    "    (7)  Z H O X P I E A F Y Q U M L B J T S N R D C V G    0.7409421121970484\n",
    "    (5)  Z H O X P I E A F Y Q U M L B J T S N R C D V G    0.7409412796324509\n",
    "    12 of 12 layouts tied for first place"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Assign letters W and K and test left/right swap\n",
    "\n",
    "Test to see if equal or higher scores are obtained for the following:\n",
    "\n",
    "    1. Assign W and K to keys 112 and 113\n",
    "    2. Swap left and right sides"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "layouts_26letters = [\n",
    "    ['Z','H','O','X','P','I','E','A','F','Y','Q','U', 'M','D','B','J','T','S','N','R','C','L','V','G', '-','-','-', '-','-','-', 'W','K'],\n",
    "    ['Z','H','O','X','P','I','E','A','F','Y','Q','U', 'M','D','B','J','T','S','N','R','C','L','V','G', '-','-','-', '-','-','-', 'K','W'],\n",
    "    ['J','B','D','M','R','N','S','T','G','V','L','C', 'X','O','H','Z','A','E','I','P','U','Q','Y','F', '-','-','-', '-','-','-', 'W','K'],\n",
    "    ['J','B','D','M','R','N','S','T','G','V','L','C', 'X','O','H','Z','A','E','I','P','U','Q','Y','F', '-','-','-', '-','-','-', 'K','W']]\n",
    "data_matrix = Factors32x32\n",
    "scores_26letters = []\n",
    "for layout_26letters in layouts_26letters:\n",
    "    scores_26letters.append(score_layout(data_matrix, layout_26letters, bigrams, bigram_frequencies, verbose=False))\n",
    "\n",
    "scores_26letters_sorted, ranks_26letters_sorted, Isort_26letters = rank_within_epsilon(scores_26letters, \n",
    "                                                                        factor32, factor=True, verbose=False)\n",
    "print('\\n    Rank                                                                   Score')\n",
    "for i, rank in enumerate(ranks_26letters_sorted):\n",
    "    layout_string = layouts_26letters[Isort_26letters[i]]\n",
    "    layout = ' '.join(layout_string)\n",
    "    print('    {0}:  {1}    {2}'.format(rank, layout, scores_26letters_sorted[i]))\n",
    "    \n",
    "print('')\n",
    "print_layout24(layouts_26letters[0])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "W above K received the highest score, but all four tied for first place:\n",
    "\n",
    "    Rank                                                                   Score\n",
    "    1:  Z H O X P I E A F Y Q U M D B J T S N R C L V G - - - - - - W K    0.6276635389970967\n",
    "    1:  J B D M R N S T G V L C X O H Z A E I P U Q Y F - - - - - - W K    0.6276633562753785\n",
    "    1:  Z H O X P I E A F Y Q U M D B J T S N R C L V G - - - - - - K W    0.627662962273589\n",
    "    1:  J B D M R N S T G V L C X O H Z A E I P U Q Y F - - - - - - K W    0.6276628500567929\n",
    "\n",
    "We will choose the top layout with vowels on the left as our engram-es layout:\n",
    "\n",
    "    Z H O X  M D B J\n",
    "    P I E A  T S N R\n",
    "    F Y Q U  C L V G\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "winner24 = ['Z','H','O','X','P','I','E','A','F','Y','Q','U', 'M','D','B','J','T','S','N','R','C','L','V','G']\n",
    "winner32 = ['Z','H','O','X','P','I','E','A','F','Y','Q','U', 'M','D','B','J','T','S','N','R','C','L','V','G', '-','-','-', '-','-','-', 'W','K']"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Test 3. Stability test: search for higher-scoring layouts by rearranging letters\n",
    "\n",
    "The following test is to see if allowing random sets of eight letters to rearrange in every possible combination improves the score of the winning layout. After randomly selecting eight letters from the top-scoring layout, creating layouts from every permutation of these letters, and computing their scores, we get identical results as the original layout. We repeated this test over a thousand times (40,320,000 layouts).\n",
    "\n",
    "The eight letters are selected from all but the 4 vowels assigned to the left side in the top-scoring layout and the 4 consonants that were assigned to the same keys in all 16 top layouts:\n",
    "\n",
    "        - - O - - I E A - - - - M - - - - S N R - - - -"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "run_stability_test = True\n",
    "if run_stability_test:\n",
    "    original_score = score_layout(Factors24x24, winner24, bigrams, bigram_frequencies, verbose=False)\n",
    "    top_score_test3 = original_score\n",
    "    nunber_of_tests = 1000\n",
    "    size_random_set = 8\n",
    "    indices = [0,1,3, 4, 8,9,10,11, 13,14,15, 16, 20,21,22,23]\n",
    "\n",
    "    #  Z H O X  M D B J\n",
    "    #  P I E A  T S N R\n",
    "    #  F Y Q U  C L V G\n",
    "\n",
    "    #  - - O -  M - - - \n",
    "    #  - I E A  - S N R\n",
    "    #  - - - -  - - - -\n",
    "\n",
    "    print(original_score)\n",
    "\n",
    "    for i in range(nunber_of_tests):\n",
    "\n",
    "        print(i)\n",
    "        \n",
    "        letters_copy = winner24.copy() \n",
    "        random_indices = []\n",
    "        while np.size(random_indices) < size_random_set:\n",
    "            random_index = indices[np.int( np.round( (np.size(indices) - 1) * np.random.random(1) )[0])]\n",
    "            if random_index not in random_indices:\n",
    "                random_indices.append(random_index)   \n",
    "        for irand in random_indices:\n",
    "            letters_copy[np.int(irand)] = ''\n",
    "\n",
    "        top_permutation_test3, top_score_test3 = permute_optimize(winner24, letters_copy, letters24, keys24, \n",
    "                                                                  Factors24x24, bigrams, bigram_frequencies,                                                                   \n",
    "                                                                  min_score=top_score_test3, verbose=False)\n",
    "\n",
    "        if ''.join(top_permutation_test3) != ''.join(winner24) and top_score_test3 > original_score:\n",
    "            print(top_score_test3)\n",
    "            print(*top_permutation_test3)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "sdl3lLOfDzT8"
   },
   "source": [
    "## Step 5: Arrange non-letter characters in easy-to-remember places <a name=\"step5\">\n",
    "    \n",
    "Now that we have all 26 letters accounted for, we turn our attention to non-letter characters, taking into account frequency of punctuation and ease of recall.\n",
    "  \n",
    "### Add diacritical marks\n",
    "\n",
    "A special diacritical key (denoted by ☆ U+2606) when simultaneously pressed with a letter adds a diacritical mark to that letter:\n",
    "  \n",
    "    ☆ + aeiouAEIOU = áéíóúÁÉÍÓÚ (acute accent)\n",
    "    ☆ + nN = ñÑ\n",
    "    ☆ + cC = çÇ\n",
    "    ☆ + Shift + [letter] = [letter] with a diaresis/umlaut: ü\n",
    "    \n",
    "For generalizability beyond Spanish:\n",
    "    \n",
    "    ☆         + AltGr + [letter] = [letter] with a grave accent: è\n",
    "    ☆ + Shift + AltGr + [letter] = [letter] with a circumflex: â\n",
    " \n",
    "\n",
    "### Add punctuation keys and number keys\n",
    "\n",
    "We will assign some of the most frequent punctuation to the six keys in the middle two columns:\n",
    "\n",
    "            Z H O X  .   \"   M D B J W\n",
    "            P I E A  ,   ☆   T S N R K\n",
    "            F Y Q U  ¿   ?   C L V G\n",
    "\n",
    "We will use the Shift key to group similar punctuation marks:\n",
    "\n",
    "            Z H O X  .:  \"'  M D B J W\n",
    "            P I E A  ,;  ☆   T S N R K\n",
    "            F Y Q U  ¿¡  ?!  C L V G\n",
    "\n",
    "**Number keys**: \n",
    "The numbers are flanked to the left and right by [square brackets], and {curly brackets} accessed by the Shift key. Each of the numbers is paired with a mathematical or logic symbol accessed by the Shift key:\n",
    "    \n",
    "    { | = ~ +   <  >   ^ & % * } \\\n",
    "    [ 1 2 3 4   5  6   7 8 9 0 ] /\n",
    "\n",
    "    1: | (vertical bar or \"pipe\" represents the logical OR operator: 1 stroke, looks like the number one)\n",
    "    2: = (equal: 2 strokes, like the Chinese character for \"2\")\n",
    "    3: ~ (tilde: \"almost equal\", often written with 3 strokes, like the Chinese character for \"3\")\n",
    "    4: + (plus: has four quadrants; resembles \"4\")\n",
    "    5 & 6: < > (\"less/greater than\"; these angle brackets are directly above the other bracket keys)\n",
    "    7: ^ (caret for logical XOR operator as well as exponentiation; resembles \"7\")\n",
    "    8: & (ampersand: logical AND operator; resembles \"8\")\n",
    "    9: % (percent: related to division; resembles \"9\")\n",
    "    0: * (asterisk: for multiplication; resembles \"0\") \n",
    "\n",
    "The three remaining keys in many common keyboards (flanking the upper right hand corner Backspace key) are displaced in special keyboards, such as the Kinesis Advantage and Ergodox. For the top right key, we will assign the forward slash and backslash: / \\\\. For one of the remaining two keys we will assign the hyphen and underscore, and to the other key, two symbols that in modern usage have significance in social media: the hash/pound sign and the \"at sign\". The hash or hashtag identifies digital content on a specific topic (the Shift key accesses the money sign). The \"at sign\" identifies a location or affiliation (such as in email addresses) and acts as a \"handle\" to identify users in popular social media platforms and online forums.\n",
    "\n",
    "The resulting engram-es layout:\n",
    "\n",
    "          [ | = ~ +   <  >   ^ & % * ] \\\n",
    "          ( 1 2 3 4   5  6   7 8 9 0 ) /\n",
    "\n",
    "            Z H O X  .:  \"'  M D B J W -_ @#\n",
    "            P I E A  ,;  ☆   T S N R K\n",
    "            F Y Q U  ¿¡  ?!  C L V G\n",
    "\n",
    "The AltGr key will access additional symbols:\n",
    "    \n",
    "    AltGr + ( = { (open curly brace)\n",
    "    AltGr + ) = } (close curly brace)\n",
    "    AltGr + 5 = « (open quote/comillas) \n",
    "    AltGr + 6 = » (close quote/comillas)\n",
    "    AltGr + - = — (em dash)\n",
    "    AltGr + ' = ` (back tick)\n",
    "    AltGr + . = • (middle dot, or \"interpunct\")\n",
    "    AltGr + s = $ (dollar currency)\n",
    "    AltGr + e = € (euro currency)\n",
    "    AltGr + l = £ (pound currency)\n",
    " \n",
    "And the extra key on ISO keyboards accesses a duplicate asterisk (*)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "colab": {
   "name": "engram-layout.ipynb",
   "provenance": []
  },
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}