Tariq - FilterTransferFromHighAccToLowAccVer02_04242019_NotWork.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "FilterTransferFromHighAccToLowAccVer02_04242019-NotWork.ipynb",
      "version": "0.3.2",
      "provenance": [],
      "collapsed_sections": [
        "kdTAcjYec3s7",
        "8rDrYIgZc20A",
        "dQQPlF6RdS2m",
        "ex0j41PLde2x",
        "Ssi32yj-d1iX",
        "U_0LHS-Xf7aN",
        "MQ5H0p_KjVCw",
        "DCCl7-vOfSZ9",
        "TEkg52f_HxEg"
      ],
      "toc_visible": true,
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/preritt/CS231n-/blob/master/Tariq%20-%20FilterTransferFromHighAccToLowAccVer02_04242019_NotWork.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "metadata": {
        "id": "kdTAcjYec3s7",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "## Load standard libraries"
      ]
    },
    {
      "metadata": {
        "id": "oCo6LRTTcxmC",
        "colab_type": "code",
        "outputId": "98861fa4-a498-41de-fe3e-99333e69ae36",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "cell_type": "code",
      "source": [
        "import keras\n",
        "from matplotlib import pyplot as plt\n",
        "import numpy as np\n",
        "import gzip\n",
        "%matplotlib inline\n",
        "from keras.models import Model\n",
        "from keras.optimizers import RMSprop\n",
        "from keras.layers import Input,Dense,Flatten,Dropout,merge,Reshape,Conv2D,MaxPooling2D,UpSampling2D,Conv2DTranspose\n",
        "from keras.layers.normalization import BatchNormalization\n",
        "from keras.models import Model,Sequential\n",
        "from keras.callbacks import ModelCheckpoint\n",
        "from keras.optimizers import Adadelta, RMSprop,SGD,Adam\n",
        "from keras import regularizers\n",
        "from keras import backend as K\n",
        "from keras.utils import to_categorical\n",
        "from keras.datasets import cifar10\n",
        "from sklearn import preprocessing\n",
        "from keras import losses\n",
        "from keras.models import model_from_json\n",
        "import keras\n",
        "from keras.models import Sequential\n",
        "from keras.utils import np_utils\n",
        "from keras.preprocessing.image import ImageDataGenerator\n",
        "from keras.layers import Dense, Activation, Flatten, Dropout, BatchNormalization\n",
        "from keras.layers import Conv2D, MaxPooling2D\n",
        "from keras.datasets import cifar10\n",
        "from keras import regularizers\n",
        "from keras.callbacks import LearningRateScheduler\n",
        "import numpy as np\n",
        "from sklearn.metrics import confusion_matrix\n",
        "from sklearn.model_selection import train_test_split\n",
        "from sklearn.utils import shuffle"
      ],
      "execution_count": 1,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Using TensorFlow backend.\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "metadata": {
        "id": "Sg_ZApCOPN9u",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "from keras.utils.vis_utils import plot_model\n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "oG3txkjndHWQ",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "keras.backend.clear_session()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "8rDrYIgZc20A",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "## Load image data"
      ]
    },
    {
      "metadata": {
        "id": "85Dlxs7adKVS",
        "colab_type": "code",
        "outputId": "211eeaee-5a0d-43e2-c187-d93712ecfcca",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 337
        }
      },
      "cell_type": "code",
      "source": [
        "from matplotlib import pyplot\n",
        "from scipy.misc import toimage\n",
        "from keras.datasets import cifar10\n",
        "def show_imgs(X):\n",
        "    pyplot.figure(1)\n",
        "    k = 0\n",
        "    for i in range(0,4):\n",
        "        for j in range(0,4):\n",
        "            pyplot.subplot2grid((4,4),(i,j))\n",
        "            pyplot.imshow(toimage(X[k]))\n",
        "            k = k+1\n",
        "    # show the plot\n",
        "    pyplot.show()\n",
        " \n",
        "(x_train, y_train), (x_test, y_test) = cifar10.load_data()\n",
        "show_imgs(x_test[:16])\n"
      ],
      "execution_count": 4,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/ipykernel_launcher.py:10: DeprecationWarning: `toimage` is deprecated!\n",
            "`toimage` is deprecated in SciPy 1.0.0, and will be removed in 1.2.0.\n",
            "Use Pillow's ``Image.fromarray`` directly instead.\n",
            "  # Remove the CWD from sys.path while we load stuff.\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAVoAAAD8CAYAAAA2Y2wxAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nS4+xitZisVh6zP8CB+zi+GJTgIQAAAAASUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 432x288 with 16 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "metadata": {
        "id": "ooaSupOjdLs9",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        ""
      ]
    },
    {
      "metadata": {
        "id": "dQQPlF6RdS2m",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "## Create Function to get data corresponding to three labels"
      ]
    },
    {
      "metadata": {
        "id": "GFyk3xzNdYNW",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "# This function takes the entire traaining data and test data and corresponding categorical label and returns the corresponding data in class indices\n",
        "def getThreeClassesTrainTest(class_indices,train_data,train_labels,test_data,test_labels):\n",
        "  all_train_data = []\n",
        "  all_train_data_label = []\n",
        "  all_test_data = []\n",
        "  all_test_data_label = []\n",
        "  for ind in class_indices:\n",
        "    this_class_location_train =  np.where(train_labels==ind)[0] # train_labels[train_labels == ind]\n",
        "    this_class_data_train = train_data[this_class_location_train,:]\n",
        "    this_class_data_train_label = train_labels[this_class_location_train]\n",
        "#     print(this_class_location_train)\n",
        "    \n",
        "    this_class_location_test = np.where(test_labels==ind)[0] #test_labels[test_labels == ind]\n",
        "    this_class_data_test = test_data[this_class_location_test,:]\n",
        "    this_class_data_test_label = test_labels[this_class_location_test]\n",
        "    \n",
        "    all_train_data.append(this_class_data_train)\n",
        "    all_test_data.append(this_class_data_test)\n",
        "    all_train_data_label.append(this_class_data_train_label)\n",
        "    all_test_data_label.append(this_class_data_test_label)\n",
        "    \n",
        "  train_data_three = all_train_data[0]\n",
        "  train_label_three = all_train_data_label[0]\n",
        "  test_data_three = all_test_data[0]\n",
        "  test_label_three = all_test_data_label[0]\n",
        "  \n",
        "  for j in range(1,len(class_indices)):\n",
        "    train_data_three = np.concatenate((train_data_three,all_train_data[j]))\n",
        "    train_label_three = np.concatenate((train_label_three,all_train_data_label[j]))\n",
        "\n",
        "    test_data_three = np.concatenate((test_data_three,all_test_data[j]))\n",
        "    test_label_three = np.concatenate((test_label_three,all_test_data_label[j]))\n",
        "#   print(train_label_three.shape)\n",
        "  train_label_three = np.reshape(train_label_three,newshape=(-1,))\n",
        "  test_label_three = np.reshape(test_label_three,newshape=(-1,))\n",
        "  return train_data_three, train_label_three, test_data_three,test_label_three"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "ex0j41PLde2x",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "## Get data for three classes using getThreeClassesTrainTest function"
      ]
    },
    {
      "metadata": {
        "id": "IpqbOTfBdhHk",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "train_three_data,train_three_label,test_three_data,test_three_label = getThreeClassesTrainTest([ 0,3,5 ],x_train,y_train,x_test,y_test)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "9gK_tiIWd2Nf",
        "colab_type": "code",
        "outputId": "6a37b812-d322-4eab-b5bd-ff3217c4173d",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 337
        }
      },
      "cell_type": "code",
      "source": [
        "plt.hist(train_three_label)"
      ],
      "execution_count": 7,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(array([5000.,    0.,    0.,    0.,    0.,    0., 5000.,    0.,    0.,\n",
              "        5000.]),\n",
              " array([0. , 0.5, 1. , 1.5, 2. , 2.5, 3. , 3.5, 4. , 4.5, 5. ]),\n",
              " <a list of 10 Patch objects>)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 7
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYAAAAD8CAYAAAB+UHOxAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n/gESZomJ0QAAAABJRU5ErkJggg==\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "metadata": {
        "id": "AEYcIMeqd4i1",
        "colab_type": "code",
        "outputId": "c60bab57-52c3-4629-cb68-5caf1a0dcd54",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 337
        }
      },
      "cell_type": "code",
      "source": [
        "plt.hist(test_three_label)"
      ],
      "execution_count": 8,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(array([1000.,    0.,    0.,    0.,    0.,    0., 1000.,    0.,    0.,\n",
              "        1000.]),\n",
              " array([0. , 0.5, 1. , 1.5, 2. , 2.5, 3. , 3.5, 4. , 4.5, 5. ]),\n",
              " <a list of 10 Patch objects>)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 8
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYAAAAD8CAYAAAB+UHOxAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4zLCBo\ndHRwOi8vbWF0cGxvdGxpYi5vcmcvnQurowAADfxJREFUeJzt3X+o3fV9x/Hna6b2h92MPy4hS9Jd\noaFDCptycQ5HGWZz/iiNf7Ri2doggfzjNjsHbbp/ZNs/FUZthSGExi0y0Yp1GFppFzSlCDPtjbW2\nmnZenDYJam7rj9aV0mV974/zsbtLExPv995z6v08H3C53+/n+znn+zka87zne885pqqQJPXn1ya9\nAEnSZBgASeqUAZCkThkASeqUAZCkThkASeqUAZCkThkASeqUAZCkTq2a9AJez7nnnlvT09OTXoYk\nvans37//B1U1dbJ5v9IBmJ6eZnZ2dtLLkKQ3lSTPnso8LwFJUqcMgCR1ygBIUqcMgCR1ygBIUqdO\nGoAktyc5kuQ7C8bOTrInyVPt+1ltPEluTTKX5PEkFy64zZY2/6kkW5bn4UiSTtWpPAP4Z+DyY8a2\nAw9W1UbgwbYPcAWwsX1tA26DUTCAm4DfAy4CbnotGpKkyThpAKrqa8CLxwxvBna17V3A1QvG76iR\nR4DVSdYCfwLsqaoXq+olYA+/HBVJ0hgt9ncAa6rqubb9PLCmba8DDi6Yd6iNnWhckjQhg98JXFWV\nZMn+z/JJtjG6fMS73vWuQfc1vf1LS7GkN+yZT101kfOqD5P6cw39/dle6f+sF/sM4IV2aYf2/Ugb\nPwxsWDBvfRs70fgvqaodVTVTVTNTUyf9KAtJ0iItNgC7gddeybMFuH/B+Efbq4EuBl5pl4q+AlyW\n5Kz2y9/L2pgkaUJOegkoyV3AHwLnJjnE6NU8nwLuSbIVeBa4pk1/ALgSmAN+AlwHUFUvJvl74Btt\n3t9V1bG/WJYkjdFJA1BVHz7BoU3HmVvA9Se4n9uB29/Q6iRJy8Z3AktSpwyAJHXKAEhSpwyAJHXK\nAEhSpwyAJHXKAEhSpwyAJHXKAEhSpwyAJHXKAEhSpwyAJHXKAEhSpwyAJHXKAEhSpwyAJHXKAEhS\npwyAJHXKAEhSpwyAJHXKAEhSpwyAJHXKAEhSpwyAJHXKAEhSpwyAJHXKAEhSpwyAJHXKAEhSpwyA\nJHXKAEhSpwyAJHVqUACS/FWSJ5J8J8ldSd6W5Lwk+5LMJfl8ktPb3Le2/bl2fHopHoAkaXEWHYAk\n64C/BGaq6r3AacC1wM3ALVX1buAlYGu7yVbgpTZ+S5snSZqQoZeAVgFvT7IKeAfwHHApcG87vgu4\num1vbvu045uSZOD5JUmLtOgAVNVh4B+A7zP6i/8VYD/wclUdbdMOAeva9jrgYLvt0Tb/nMWeX5I0\nzJBLQGcx+qn+POA3gTOAy4cuKMm2JLNJZufn54fenSTpBIZcAvoj4D+rar6q/hu4D7gEWN0uCQGs\nBw637cPABoB2/Ezgh8feaVXtqKqZqpqZmpoasDxJ0usZEoDvAxcneUe7lr8JeBLYC3ywzdkC3N+2\nd7d92vGHqqoGnF+SNMCQ3wHsY/TL3EeBb7f72gF8ArgxyRyja/w72012Aue08RuB7QPWLUkaaNXJ\np5xYVd0E3HTM8NPARceZ+1PgQ0POJ0laOr4TWJI6ZQAkqVMGQJI6ZQAkqVMGQJI6ZQAkqVMGQJI6\nZQAkqVMGQJI6ZQAkqVMGQJI6ZQAkqVMGQJI6ZQAkqVMGQJI6ZQAkqVMGQJI6ZQAkqVMGQJI6ZQAk\nqVMGQJI6ZQAkqVMGQJI6ZQAkqVMGQJI6ZQAkqVMGQJI6ZQAkqVMGQJI6ZQAkqVMGQJI6ZQAkqVMG\nQJI6NSgASVYnuTfJd5McSPL7Sc5OsifJU+37WW1uktyaZC7J40kuXJqHIElajKHPAD4LfLmqfhv4\nHeAAsB14sKo2Ag+2fYArgI3taxtw28BzS5IGWHQAkpwJvA/YCVBVP6uql4HNwK42bRdwddveDNxR\nI48Aq5OsXfTKJUmDDHkGcB4wD/xTkm8m+VySM4A1VfVcm/M8sKZtrwMOLrj9oTYmSZqAIQFYBVwI\n3FZVFwD/xf9d7gGgqgqoN3KnSbYlmU0yOz8/P2B5kqTXMyQAh4BDVbWv7d/LKAgvvHZpp30/0o4f\nBjYsuP36Nvb/VNWOqpqpqpmpqakBy5MkvZ5FB6CqngcOJnlPG9oEPAnsBra0sS3A/W17N/DR9mqg\ni4FXFlwqkiSN2aqBt/8L4M4kpwNPA9cxiso9SbYCzwLXtLkPAFcCc8BP2lxJ0oQMCkBVPQbMHOfQ\npuPMLeD6IeeTJC0d3wksSZ0yAJLUKQMgSZ0yAJLUKQMgSZ0yAJLUKQMgSZ0yAJLUKQMgSZ0yAJLU\nKQMgSZ0yAJLUKQMgSZ0yAJLUKQMgSZ0yAJLUKQMgSZ0yAJLUKQMgSZ0yAJLUKQMgSZ0yAJLUKQMg\nSZ0yAJLUKQMgSZ0yAJLUKQMgSZ0yAJLUKQMgSZ0yAJLUKQMgSZ0yAJLUqcEBSHJakm8m+WLbPy/J\nviRzST6f5PQ2/ta2P9eOTw89tyRp8ZbiGcANwIEF+zcDt1TVu4GXgK1tfCvwUhu/pc2TJE3IoAAk\nWQ9cBXyu7Qe4FLi3TdkFXN22N7d92vFNbb4kaQKGPgP4DPBx4Odt/xzg5ao62vYPAeva9jrgIEA7\n/kqbL0magEUHIMn7gSNVtX8J10OSbUlmk8zOz88v5V1LkhYY8gzgEuADSZ4B7mZ06eezwOokq9qc\n9cDhtn0Y2ADQjp8J/PDYO62qHVU1U1UzU1NTA5YnSXo9iw5AVX2yqtZX1TRwLfBQVf0psBf4YJu2\nBbi/be9u+7TjD1VVLfb8kqRhluN9AJ8Abkwyx+ga/842vhM4p43fCGxfhnNLkk7RqpNPObmq+irw\n1bb9NHDRceb8FPjQUpxPkjSc7wSWpE4ZAEnqlAGQpE4ZAEnqlAGQpE4ZAEnqlAGQpE4ZAEnqlAGQ\npE4ZAEnqlAGQpE4ZAEnqlAGQpE4ZAEnqlAGQpE4ZAEnqlAGQpE4ZAEnqlAGQpE4ZAEnqlAGQpE4Z\nAEnqlAGQpE4ZAEnqlAGQpE4ZAEnqlAGQpE4ZAEnqlAGQpE4ZAEnqlAGQpE4ZAEnqlAGQpE4tOgBJ\nNiTZm+TJJE8kuaGNn51kT5Kn2vez2niS3JpkLsnjSS5cqgchSXrjhjwDOAr8dVWdD1wMXJ/kfGA7\n8GBVbQQebPsAVwAb29c24LYB55YkDbToAFTVc1X1aNv+MXAAWAdsBna1abuAq9v2ZuCOGnkEWJ1k\n7aJXLkkaZEl+B5BkGrgA2Aesqarn2qHngTVtex1wcMHNDrWxY+9rW5LZJLPz8/NLsTxJ0nEMDkCS\ndwJfAD5WVT9aeKyqCqg3cn9VtaOqZqpqZmpqaujyJEknMCgASd7C6C//O6vqvjb8wmuXdtr3I238\nMLBhwc3XtzFJ0gQMeRVQgJ3Agar69IJDu4EtbXsLcP+C8Y+2VwNdDLyy4FKRJGnMVg247SXAR4Bv\nJ3msjf0N8CngniRbgWeBa9qxB4ArgTngJ8B1A84tSRpo0QGoqoeBnODwpuPML+D6xZ5PkrS0fCew\nJHXKAEhSpwyAJHXKAEhSpwyAJHXKAEhSpwyAJHXKAEhSpwyAJHXKAEhSpwyAJHXKAEhSpwyAJHXK\nAEhSpwyAJHXKAEhSpwyAJHXKAEhSpwyAJHXKAEhSpwyAJHXKAEhSpwyAJHXKAEhSpwyAJHXKAEhS\npwyAJHXKAEhSpwyAJHXKAEhSpwyAJHXKAEhSp8YegCSXJ/lekrkk28d9fknSyFgDkOQ04B+BK4Dz\ngQ8nOX+ca5AkjYz7GcBFwFxVPV1VPwPuBjaPeQ2SJMYfgHXAwQX7h9qYJGnMVk16AcdKsg3Y1nZf\nTfK9AXd3LvCD4at6Y3LzuM/4CxN5vBPmYx4j/2yPT24e9Jh/61QmjTsAh4ENC/bXt7FfqKodwI6l\nOFmS2aqaWYr7ejPo7fGCj7kXPublMe5LQN8ANiY5L8npwLXA7jGvQZLEmJ8BVNXRJH8OfAU4Dbi9\nqp4Y5xokSSNj/x1AVT0APDCm0y3JpaQ3kd4eL/iYe+FjXgapquU+hyTpV5AfBSFJnVqRAejt4yaS\n3J7kSJLvTHot45JkQ5K9SZ5M8kSSGya9puWW5G1Jvp7kW+0x/+2k1zQOSU5L8s0kX5z0WsYlyTNJ\nvp3ksSSzy3aelXYJqH3cxH8Af8zojWbfAD5cVU9OdGHLKMn7gFeBO6rqvZNezzgkWQusrapHk/w6\nsB+4eoX/ew5wRlW9muQtwMPADVX1yISXtqyS3AjMAL9RVe+f9HrGIckzwExVLet7H1biM4DuPm6i\nqr4GvDjpdYxTVT1XVY+27R8DB1jh7yqvkVfb7lva18r6Ce4YSdYDVwGfm/RaVqKVGAA/bqIzSaaB\nC4B9k13J8muXQx4DjgB7qmqlP+bPAB8Hfj7phYxZAf+WZH/7dIRlsRIDoI4keSfwBeBjVfWjSa9n\nuVXV/1TV7zJ6F/1FSVbsJb8k7weOVNX+Sa9lAv6gqi5k9MnJ17fLvEtuJQbgpB83oZWhXQf/AnBn\nVd036fWMU1W9DOwFLp/0WpbRJcAH2vXwu4FLk/zLZJc0HlV1uH0/Avwro0vbS24lBsCPm+hA+4Xo\nTuBAVX160usZhyRTSVa37bczeqHDdye7quVTVZ+sqvVVNc3ov+OHqurPJrysZZfkjPbCBpKcAVwG\nLMsr/FZcAKrqKPDax00cAO5Z6R83keQu4N+B9yQ5lGTrpNc0BpcAH2H0U+Fj7evKSS9qma0F9iZ5\nnNEPOnuqqpuXRnZkDfBwkm8BXwe+VFVfXo4TrbiXgUqSTs2KewYgSTo1BkCSOmUAJKlTBkCSOmUA\nJKlTBkCSOmUAJKlTBkCSOvW/lK9vO4tP9cMAAAAASUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "metadata": {
        "id": "z4n_WiMNd_Hf",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "train_three_data, train_three_label = shuffle(train_three_data, train_three_label )\n",
        "test_three_data,test_three_label = shuffle(test_three_data,test_three_label )"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "VT_AKqQmeBk_",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "le = preprocessing.LabelEncoder()\n",
        "le.fit_transform(train_three_label)\n",
        "train_three_label_cat = le.transform(train_three_label)\n",
        "test_three_label_cat = le.transform(test_three_label)\n",
        "train_three_label_one_hot = to_categorical(train_three_label_cat)\n",
        "test_three_label_one_hot = to_categorical(test_three_label_cat)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "rWEjK_exeEaJ",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "# split train data and to create validation data\n",
        "train_X,valid_X,train_ground,valid_ground = train_test_split(train_three_data,\n",
        "                                                             train_three_label_one_hot,\n",
        "                                                             test_size=0.2,\n",
        "                                                             random_state=13)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "mOFKlZWKeH25",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "num_classes = test_three_label_one_hot.shape[1]"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "Ssi32yj-d1iX",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "## Define the model"
      ]
    },
    {
      "metadata": {
        "id": "uUISHV2weN0t",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "x, y = 32, 32\n",
        "inChannel = 3\n",
        "input_img = Input(shape = (x, y, inChannel))\n",
        "ep = 10 \n",
        "weight_decay = 1e-4"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "Ich0RbYnebwW",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "def cifar10Classifier(input_img):\n",
        "  \n",
        "  conv1 = (Conv2D(32, (3,3), padding='same', kernel_regularizer=regularizers.l2(weight_decay), input_shape=x_train.shape[1:], activation='elu'))(input_img)\n",
        "  conv1 = BatchNormalization()(conv1)\n",
        "  conv1 = (Conv2D(32, (3,3), padding='same', activation= 'elu',kernel_regularizer=regularizers.l2(weight_decay)))(conv1)\n",
        "  conv1 = BatchNormalization()(conv1)\n",
        "  pool1 = MaxPooling2D(pool_size=(2, 2))(conv1) #14 x 14 x 32\n",
        "  drop1= Dropout(.2)(pool1)\n",
        "\n",
        "  conv2 = (Conv2D(64, (3,3), padding='same', kernel_regularizer=regularizers.l2(weight_decay), activation='elu'))(drop1)\n",
        "  conv2 = BatchNormalization()(conv2)\n",
        "  conv2 = (Conv2D(64, (3,3), padding='same', activation= 'elu',kernel_regularizer=regularizers.l2(weight_decay)))(conv2)\n",
        "  conv2 = BatchNormalization()(conv2)\n",
        "  pool2 = MaxPooling2D(pool_size=(2, 2))(conv2) #14 x 14 x 32\n",
        "  drop2= Dropout(.3)(pool2)\n",
        "\n",
        "  conv3 = (Conv2D(128, (3,3), padding='same', kernel_regularizer=regularizers.l2(weight_decay), activation='elu'))(drop2)\n",
        "  conv3 = BatchNormalization()(conv3)\n",
        "  conv3 = (Conv2D(128, (3,3), padding='same', activation= 'elu',kernel_regularizer=regularizers.l2(weight_decay)))(conv3)\n",
        "  conv3 = BatchNormalization()(conv3)\n",
        "  pool3 = MaxPooling2D(pool_size=(2, 2))(conv3) #14 x 14 x 32\n",
        "  drop3= Dropout(.4)(pool3)\n",
        "\n",
        "  flat = Flatten()(drop3)\n",
        "  out = Dense(num_classes, activation='softmax')(flat)\n",
        "  \n",
        "  return out"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "U_0LHS-Xf7aN",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "## Define functions for calculation of DI score\n",
        "getDIScoreOfLayer(layer_output, true_labels), \n",
        "getDIScoreOfLayerCNNLayer(cnn_outputs, true_labels)\n"
      ]
    },
    {
      "metadata": {
        "id": "xsLk-f7sf6zd",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "def getDIScoreOfLayer(layer_output, true_labels):\n",
        "  # computeScatterMatrix\n",
        "  rho = 1e-6\n",
        "  mu = layer_output.mean(axis=0)\n",
        "#   print(mu.shape)\n",
        "  mu = mu.reshape((1,-1))\n",
        "  centered_data = layer_output - mu\n",
        "  S_bar = np.matmul(centered_data.T,centered_data)\n",
        "  # compute between class matrix\n",
        "  total_classes = true_labels.max()\n",
        "  S_B = np.zeros(shape = (mu.shape[1],mu.shape[1]))\n",
        "  S_W = np.zeros(shape = (mu.shape[1],mu.shape[1]))\n",
        "  unique_classes = np.unique(true_labels)\n",
        "  unique_classes_list = np.ndarray.tolist(unique_classes)\n",
        "  for l in unique_classes_list:\n",
        "    class_l_index = np.where(true_labels == l)\n",
        "    class_l_output = layer_output[class_l_index[0],:]\n",
        "    mu_l = class_l_output.mean(axis=0)\n",
        "    mu_l = mu_l.reshape((1,-1))\n",
        "    total_examples_this_class = class_l_index[0].shape   \n",
        "    ss = mu_l - mu\n",
        "    S_B = S_B + total_examples_this_class*np.matmul((mu_l - mu).T,(mu_l - mu))\n",
        "\n",
        "  DI_Score = np.trace(np.matmul(np.linalg.inv(S_bar + rho*np.eye(S_bar.shape[0])),S_B))\n",
        "  return DI_Score"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "ubrpaIpFgBzn",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "def getDIScoreOfLayerCNNLayer(cnn_outputs, true_labels):\n",
        "  # computeScatterMatrix\n",
        "  number_of_filters = cnn_outputs.shape[3]\n",
        "  cnn_di_scores = []\n",
        "  for filt in range(0,number_of_filters):\n",
        "    outputs_filter = cnn_outputs[:,:,:,filt]\n",
        "    outputs_filter_reshaped = np.reshape(outputs_filter, newshape=(cnn_outputs.shape[0],-1))\n",
        "    this_filter_score = getDIScoreOfLayer(outputs_filter_reshaped, true_labels)\n",
        "    cnn_di_scores.append(this_filter_score)\n",
        "  return cnn_di_scores"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "MQ5H0p_KjVCw",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "## define a function to get and plot di score of a layer. It takes inputs - trained model, layer name to get DI scores and returns a plot with some statistics"
      ]
    },
    {
      "metadata": {
        "id": "o9MHI1LFjUZs",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "def createCNNLayerDIScorePlot(trained_model, layer_name):\n",
        "  get_cnn1_output_function = K.function([trained_model.layers[0].input],\n",
        "                                  [trained_model.get_layer(layer_name).output])   #get_layer test_three_label\n",
        "  cnn1_output = get_cnn1_output_function([train_three_data])[0]\n",
        "  di_scores = getDIScoreOfLayerCNNLayer(cnn1_output, train_three_label)\n",
        "  # plot DI score\n",
        "  y_pos  = np.arange(len(di_scores))\n",
        "  plt.bar(y_pos, di_scores, align='center', alpha=0.5)\n",
        "  plt.xlabel('FilterNumber')\n",
        "  plt.ylabel('DI Score')\n",
        "\n",
        "  print('Mean of DI SCore: ',np.mean(di_scores))\n",
        "  print('STD of DI Score: ', np.std(di_scores))\n",
        "  print('MAX of DI Score: ', np.max(di_scores))\n",
        "  plt.title('DI score - Data - D_L, Network -N_L, Conv_6')\n",
        "  plt.ylim([0,1])\n",
        "  return di_scores"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "DCCl7-vOfSZ9",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "## load pretrained model"
      ]
    },
    {
      "metadata": {
        "id": "jqIkmDGkfRtD",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "loaded_model_low_accuracy = keras.models.load_model('lowAccuracyModelData.h5')\n",
        "loaded_model_high_accuracy = keras.models.load_model('highAccuracyModelData.h5')\n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "z7J98O5qfi8Z",
        "colab_type": "code",
        "outputId": "d67810ce-1ae8-409f-c670-4fa7b4e34b3b",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "cell_type": "code",
      "source": [
        "loaded_model_low_accuracy.evaluate(valid_X, valid_ground, verbose=0)"
      ],
      "execution_count": 19,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[0.31193556666374206, 0.9073333331743876]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 19
        }
      ]
    },
    {
      "metadata": {
        "id": "tzbIXvcEoNMj",
        "colab_type": "code",
        "outputId": "22f16fa4-dee6-4fca-97a8-ec6abfc3c784",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 850
        }
      },
      "cell_type": "code",
      "source": [
        "loaded_model_low_accuracy.summary()"
      ],
      "execution_count": 20,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "input_1 (InputLayer)         (None, 32, 32, 3)         0         \n",
            "_________________________________________________________________\n",
            "conv2d_1 (Conv2D)            (None, 32, 32, 32)        896       \n",
            "_________________________________________________________________\n",
            "batch_normalization_1 (Batch (None, 32, 32, 32)        128       \n",
            "_________________________________________________________________\n",
            "conv2d_2 (Conv2D)            (None, 32, 32, 32)        9248      \n",
            "_________________________________________________________________\n",
            "batch_normalization_2 (Batch (None, 32, 32, 32)        128       \n",
            "_________________________________________________________________\n",
            "max_pooling2d_1 (MaxPooling2 (None, 16, 16, 32)        0         \n",
            "_________________________________________________________________\n",
            "dropout_1 (Dropout)          (None, 16, 16, 32)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_3 (Conv2D)            (None, 16, 16, 64)        18496     \n",
            "_________________________________________________________________\n",
            "batch_normalization_3 (Batch (None, 16, 16, 64)        256       \n",
            "_________________________________________________________________\n",
            "conv2d_4 (Conv2D)            (None, 16, 16, 64)        36928     \n",
            "_________________________________________________________________\n",
            "batch_normalization_4 (Batch (None, 16, 16, 64)        256       \n",
            "_________________________________________________________________\n",
            "max_pooling2d_2 (MaxPooling2 (None, 8, 8, 64)          0         \n",
            "_________________________________________________________________\n",
            "dropout_2 (Dropout)          (None, 8, 8, 64)          0         \n",
            "_________________________________________________________________\n",
            "conv2d_5 (Conv2D)            (None, 8, 8, 128)         73856     \n",
            "_________________________________________________________________\n",
            "batch_normalization_5 (Batch (None, 8, 8, 128)         512       \n",
            "_________________________________________________________________\n",
            "conv2d_6 (Conv2D)            (None, 8, 8, 128)         147584    \n",
            "_________________________________________________________________\n",
            "batch_normalization_6 (Batch (None, 8, 8, 128)         512       \n",
            "_________________________________________________________________\n",
            "max_pooling2d_3 (MaxPooling2 (None, 4, 4, 128)         0         \n",
            "_________________________________________________________________\n",
            "dropout_3 (Dropout)          (None, 4, 4, 128)         0         \n",
            "_________________________________________________________________\n",
            "flatten_1 (Flatten)          (None, 2048)              0         \n",
            "_________________________________________________________________\n",
            "dense_1 (Dense)              (None, 3)                 6147      \n",
            "=================================================================\n",
            "Total params: 294,947\n",
            "Trainable params: 294,051\n",
            "Non-trainable params: 896\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "UijDrRL2khSY",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "## Create a plot for DI score using function createCNNLayerDIScorePlot()"
      ]
    },
    {
      "metadata": {
        "id": "zQYh3TElhl5R",
        "colab_type": "code",
        "outputId": "85180a8e-22bd-4920-c116-3b960065352a",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 346
        }
      },
      "cell_type": "code",
      "source": [
        "# now get DI score of a layer from high accuracy and low accuracy model\n",
        "layer_name = 'conv2d_2'\n",
        "discores_low_acc_model = createCNNLayerDIScorePlot(loaded_model_low_accuracy, layer_name)"
      ],
      "execution_count": 21,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Mean of DI SCore:  0.5925375772649611\n",
            "STD of DI Score:  0.056699249107349546\n",
            "MAX of DI Score:  0.6699787896299502\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYUAAAEWCAYAAACJ0YulAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nkqSOoSBJ6hgKkqSOoSBJ6vx/AQgiF0LFN4gAAAAASUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "metadata": {
        "id": "7B0fFH7NiOh_",
        "colab_type": "code",
        "outputId": "efc9571d-db14-4bda-ea43-b411e9ba424b",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 346
        }
      },
      "cell_type": "code",
      "source": [
        "# now get DI score of a layer from high accuracy and low accuracy model\n",
        "layer_name = 'conv2d_2'\n",
        "\n",
        "discores_low_acc_model = createCNNLayerDIScorePlot(loaded_model_high_accuracy, layer_name)"
      ],
      "execution_count": 22,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Mean of DI SCore:  0.6055427452303304\n",
            "STD of DI Score:  0.04535087828139873\n",
            "MAX of DI Score:  0.7097921003254144\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYUAAAEWCAYAAACJ0YulAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "metadata": {
        "id": "Sk_0a6JlFux5",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "# returns \n",
        "def highAccuracyModelTillConv2(input_img, channel_indices_second_layer): \n",
        "  conv1_1 = (Conv2D(32, (3,3), padding='same', kernel_regularizer=regularizers.l2(weight_decay), input_shape=x_train.shape[1:], activation='elu', name = 'HighAccuracyConv1'))(input_img)\n",
        "  conv1_2 = BatchNormalization()(conv1_1)\n",
        "  conv2_1 = (Conv2D(32, (3,3), padding='same', activation= 'elu', name = 'HighAccuracyConv2',kernel_regularizer=regularizers.l2(weight_decay)))(conv1_2)\n",
        "  conv2_2 = BatchNormalization()(conv2_1)  \n",
        "  return conv2_1"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "5cCMzZdBFuvN",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "import tensorflow as tf\n",
        "def cifar10ClassifierTransfer(input_img, conv2_high):\n",
        "  \n",
        "  # Add additional inputs to \n",
        "  \n",
        "  conv1_1 = (Conv2D(32, (3,3), padding='same', kernel_regularizer=regularizers.l2(weight_decay), input_shape=x_train.shape[1:], activation='elu'))(input_img)\n",
        "  conv1_2 = BatchNormalization()(conv1_1)\n",
        "  conv2_1 = (Conv2D(32, (3,3), padding='same', activation= 'elu',kernel_regularizer=regularizers.l2(weight_decay)))(conv1_2)\n",
        "  \n",
        "  conv2_high = tf.gather(conv2_high,[0, 1, 3], axis=3)\n",
        "  \n",
        "  concat_layer = keras.layers.merge.Concatenate(axis=3)([conv2_1,conv2_high])\n",
        "  conv2_2 = BatchNormalization()(concat_layer)\n",
        "  pool2_3 = MaxPooling2D(pool_size=(2, 2))(conv2_2) #14 x 14 x 32\n",
        "  drop2_4= Dropout(.2)(pool2_3)\n",
        "\n",
        "  conv3_1 = (Conv2D(64, (3,3), padding='same', kernel_regularizer=regularizers.l2(weight_decay), activation='elu'))(drop2_4)\n",
        "  conv3_2 = BatchNormalization()(conv3_1)\n",
        "  conv4_1 = (Conv2D(64, (3,3), padding='same', activation= 'elu',kernel_regularizer=regularizers.l2(weight_decay)))(conv3_2)\n",
        "  conv4_2 = BatchNormalization()(conv4_1)\n",
        "  pool4_3 = MaxPooling2D(pool_size=(2, 2))(conv4_2) #14 x 14 x 32\n",
        "  drop4_4= Dropout(.3)(pool4_3)\n",
        "\n",
        "  conv5_1 = (Conv2D(128, (3,3), padding='same', kernel_regularizer=regularizers.l2(weight_decay), activation='elu'))(drop4_4)\n",
        "  conv5_2 = BatchNormalization()(conv5_1)\n",
        "  conv6_1 = (Conv2D(128, (3,3), padding='same', activation= 'elu',kernel_regularizer=regularizers.l2(weight_decay)))(conv5_2)\n",
        "  conv6_2 = BatchNormalization()(conv6_1)\n",
        "  pool6_3 = MaxPooling2D(pool_size=(2, 2))(conv6_2) #14 x 14 x 32\n",
        "  drop6_4= Dropout(.4)(pool6_3)\n",
        "\n",
        "  flat = Flatten()(drop6_4)\n",
        "  out = Dense(num_classes, activation='softmax')(flat)\n",
        "  \n",
        "  return out                    "
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "fo4MEEIwMpYD",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "# input_img"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "fzUmKtVGFusU",
        "colab_type": "code",
        "outputId": "fb5b44f6-8946-457e-a010-242eca45ef7d",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 394
        }
      },
      "cell_type": "code",
      "source": [
        "channel_indices_second_layer = [1,2,3,4,5]\n",
        "high_acc_output = highAccuracyModelTillConv2(input_img, channel_indices_second_layer)\n",
        "modelFilt = Model(input_img, \n",
        "              cifar10ClassifierTransfer(input_img,\n",
        "              highAccuracyModelTillConv2(input_img, channel_indices_second_layer)))\n",
        "#data augmentation\n",
        "datagen = ImageDataGenerator(\n",
        "    rotation_range=15,\n",
        "    width_shift_range=0.1,\n",
        "    height_shift_range=0.1,\n",
        "    horizontal_flip=True,\n",
        "    )\n",
        "datagen.fit(train_three_data)\n",
        " \n",
        "#training\n",
        "batch_size = 64\n",
        " \n",
        "opt_rms = keras.optimizers.rmsprop(lr=0.001,decay=1e-6)\n",
        "# model.compile(loss='categorical_crossentropy', optimizer=opt_rms, metrics=['accuracy'])\n",
        "\n",
        "modelFilt.compile(loss=losses.categorical_crossentropy, optimizer=opt_rms, metrics=['accuracy'])\n",
        "\n",
        "modelFilt.summary()\n",
        " "
      ],
      "execution_count": 176,
      "outputs": [
        {
          "output_type": "error",
          "ename": "AttributeError",
          "evalue": "ignored",
          "traceback": [
            "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
            "\u001b[0;31mAttributeError\u001b[0m                            Traceback (most recent call last)",
            "\u001b[0;32m<ipython-input-176-48a6a3044e0f>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[1;32m      3\u001b[0m modelFilt = Model(input_img, \n\u001b[1;32m      4\u001b[0m               cifar10ClassifierTransfer(input_img,\n\u001b[0;32m----> 5\u001b[0;31m               highAccuracyModelTillConv2(input_img, channel_indices_second_layer)))\n\u001b[0m\u001b[1;32m      6\u001b[0m \u001b[0;31m#data augmentation\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      7\u001b[0m datagen = ImageDataGenerator(\n",
            "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/keras/legacy/interfaces.py\u001b[0m in \u001b[0;36mwrapper\u001b[0;34m(*args, **kwargs)\u001b[0m\n\u001b[1;32m     89\u001b[0m                 warnings.warn('Update your `' + object_name + '` call to the ' +\n\u001b[1;32m     90\u001b[0m                               'Keras 2 API: ' + signature, stacklevel=2)\n\u001b[0;32m---> 91\u001b[0;31m             \u001b[0;32mreturn\u001b[0m \u001b[0mfunc\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     92\u001b[0m         \u001b[0mwrapper\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_original_function\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mfunc\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     93\u001b[0m         \u001b[0;32mreturn\u001b[0m \u001b[0mwrapper\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/keras/engine/network.py\u001b[0m in \u001b[0;36m__init__\u001b[0;34m(self, *args, **kwargs)\u001b[0m\n\u001b[1;32m     91\u001b[0m                 'inputs' in kwargs and 'outputs' in kwargs):\n\u001b[1;32m     92\u001b[0m             \u001b[0;31m# Graph network\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 93\u001b[0;31m             \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_init_graph_network\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     94\u001b[0m         \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     95\u001b[0m             \u001b[0;31m# Subclassed network\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/keras/engine/network.py\u001b[0m in \u001b[0;36m_init_graph_network\u001b[0;34m(self, inputs, outputs, name)\u001b[0m\n\u001b[1;32m    229\u001b[0m         \u001b[0;31m# Keep track of the network's nodes and layers.\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    230\u001b[0m         nodes, nodes_by_depth, layers, layers_by_depth = _map_graph_network(\n\u001b[0;32m--> 231\u001b[0;31m             self.inputs, self.outputs)\n\u001b[0m\u001b[1;32m    232\u001b[0m         \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_network_nodes\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mnodes\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    233\u001b[0m         \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_nodes_by_depth\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mnodes_by_depth\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/keras/engine/network.py\u001b[0m in \u001b[0;36m_map_graph_network\u001b[0;34m(inputs, outputs)\u001b[0m\n\u001b[1;32m   1364\u001b[0m                   \u001b[0mlayer\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mlayer\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1365\u001b[0m                   \u001b[0mnode_index\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mnode_index\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1366\u001b[0;31m                   tensor_index=tensor_index)\n\u001b[0m\u001b[1;32m   1367\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1368\u001b[0m     \u001b[0;32mfor\u001b[0m \u001b[0mnode\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mreversed\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mnodes_in_decreasing_depth\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/keras/engine/network.py\u001b[0m in \u001b[0;36mbuild_map\u001b[0;34m(tensor, finished_nodes, nodes_in_progress, layer, node_index, tensor_index)\u001b[0m\n\u001b[1;32m   1351\u001b[0m             \u001b[0mtensor_index\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mnode\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtensor_indices\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mi\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1352\u001b[0m             build_map(x, finished_nodes, nodes_in_progress, layer,\n\u001b[0;32m-> 1353\u001b[0;31m                       node_index, tensor_index)\n\u001b[0m\u001b[1;32m   1354\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1355\u001b[0m         \u001b[0mfinished_nodes\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0madd\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mnode\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/keras/engine/network.py\u001b[0m in \u001b[0;36mbuild_map\u001b[0;34m(tensor, finished_nodes, nodes_in_progress, layer, node_index, tensor_index)\u001b[0m\n\u001b[1;32m   1351\u001b[0m             \u001b[0mtensor_index\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mnode\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtensor_indices\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mi\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1352\u001b[0m             build_map(x, finished_nodes, nodes_in_progress, layer,\n\u001b[0;32m-> 1353\u001b[0;31m                       node_index, tensor_index)\n\u001b[0m\u001b[1;32m   1354\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1355\u001b[0m         \u001b[0mfinished_nodes\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0madd\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mnode\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/keras/engine/network.py\u001b[0m in \u001b[0;36mbuild_map\u001b[0;34m(tensor, finished_nodes, nodes_in_progress, layer, node_index, tensor_index)\u001b[0m\n\u001b[1;32m   1351\u001b[0m             \u001b[0mtensor_index\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mnode\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtensor_indices\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mi\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1352\u001b[0m             build_map(x, finished_nodes, nodes_in_progress, layer,\n\u001b[0;32m-> 1353\u001b[0;31m                       node_index, tensor_index)\n\u001b[0m\u001b[1;32m   1354\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1355\u001b[0m         \u001b[0mfinished_nodes\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0madd\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mnode\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/keras/engine/network.py\u001b[0m in \u001b[0;36mbuild_map\u001b[0;34m(tensor, finished_nodes, nodes_in_progress, layer, node_index, tensor_index)\u001b[0m\n\u001b[1;32m   1351\u001b[0m             \u001b[0mtensor_index\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mnode\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtensor_indices\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mi\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1352\u001b[0m             build_map(x, finished_nodes, nodes_in_progress, layer,\n\u001b[0;32m-> 1353\u001b[0;31m                       node_index, tensor_index)\n\u001b[0m\u001b[1;32m   1354\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1355\u001b[0m         \u001b[0mfinished_nodes\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0madd\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mnode\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/keras/engine/network.py\u001b[0m in \u001b[0;36mbuild_map\u001b[0;34m(tensor, finished_nodes, nodes_in_progress, layer, node_index, tensor_index)\u001b[0m\n\u001b[1;32m   1351\u001b[0m             \u001b[0mtensor_index\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mnode\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtensor_indices\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mi\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1352\u001b[0m             build_map(x, finished_nodes, nodes_in_progress, layer,\n\u001b[0;32m-> 1353\u001b[0;31m                       node_index, tensor_index)\n\u001b[0m\u001b[1;32m   1354\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1355\u001b[0m         \u001b[0mfinished_nodes\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0madd\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mnode\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/keras/engine/network.py\u001b[0m in \u001b[0;36mbuild_map\u001b[0;34m(tensor, finished_nodes, nodes_in_progress, layer, node_index, tensor_index)\u001b[0m\n\u001b[1;32m   1351\u001b[0m             \u001b[0mtensor_index\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mnode\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtensor_indices\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mi\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1352\u001b[0m             build_map(x, finished_nodes, nodes_in_progress, layer,\n\u001b[0;32m-> 1353\u001b[0;31m                       node_index, tensor_index)\n\u001b[0m\u001b[1;32m   1354\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1355\u001b[0m         \u001b[0mfinished_nodes\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0madd\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mnode\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/keras/engine/network.py\u001b[0m in \u001b[0;36mbuild_map\u001b[0;34m(tensor, finished_nodes, nodes_in_progress, layer, node_index, tensor_index)\u001b[0m\n\u001b[1;32m   1351\u001b[0m             \u001b[0mtensor_index\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mnode\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtensor_indices\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mi\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1352\u001b[0m             build_map(x, finished_nodes, nodes_in_progress, layer,\n\u001b[0;32m-> 1353\u001b[0;31m                       node_index, tensor_index)\n\u001b[0m\u001b[1;32m   1354\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1355\u001b[0m         \u001b[0mfinished_nodes\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0madd\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mnode\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/keras/engine/network.py\u001b[0m in \u001b[0;36mbuild_map\u001b[0;34m(tensor, finished_nodes, nodes_in_progress, layer, node_index, tensor_index)\u001b[0m\n\u001b[1;32m   1351\u001b[0m             \u001b[0mtensor_index\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mnode\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtensor_indices\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mi\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1352\u001b[0m             build_map(x, finished_nodes, nodes_in_progress, layer,\n\u001b[0;32m-> 1353\u001b[0;31m                       node_index, tensor_index)\n\u001b[0m\u001b[1;32m   1354\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1355\u001b[0m         \u001b[0mfinished_nodes\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0madd\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mnode\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/keras/engine/network.py\u001b[0m in \u001b[0;36mbuild_map\u001b[0;34m(tensor, finished_nodes, nodes_in_progress, layer, node_index, tensor_index)\u001b[0m\n\u001b[1;32m   1351\u001b[0m             \u001b[0mtensor_index\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mnode\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtensor_indices\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mi\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1352\u001b[0m             build_map(x, finished_nodes, nodes_in_progress, layer,\n\u001b[0;32m-> 1353\u001b[0;31m                       node_index, tensor_index)\n\u001b[0m\u001b[1;32m   1354\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1355\u001b[0m         \u001b[0mfinished_nodes\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0madd\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mnode\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/keras/engine/network.py\u001b[0m in \u001b[0;36mbuild_map\u001b[0;34m(tensor, finished_nodes, nodes_in_progress, layer, node_index, tensor_index)\u001b[0m\n\u001b[1;32m   1351\u001b[0m             \u001b[0mtensor_index\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mnode\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtensor_indices\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mi\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1352\u001b[0m             build_map(x, finished_nodes, nodes_in_progress, layer,\n\u001b[0;32m-> 1353\u001b[0;31m                       node_index, tensor_index)\n\u001b[0m\u001b[1;32m   1354\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1355\u001b[0m         \u001b[0mfinished_nodes\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0madd\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mnode\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/keras/engine/network.py\u001b[0m in \u001b[0;36mbuild_map\u001b[0;34m(tensor, finished_nodes, nodes_in_progress, layer, node_index, tensor_index)\u001b[0m\n\u001b[1;32m   1351\u001b[0m             \u001b[0mtensor_index\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mnode\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtensor_indices\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mi\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1352\u001b[0m             build_map(x, finished_nodes, nodes_in_progress, layer,\n\u001b[0;32m-> 1353\u001b[0;31m                       node_index, tensor_index)\n\u001b[0m\u001b[1;32m   1354\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1355\u001b[0m         \u001b[0mfinished_nodes\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0madd\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mnode\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/keras/engine/network.py\u001b[0m in \u001b[0;36mbuild_map\u001b[0;34m(tensor, finished_nodes, nodes_in_progress, layer, node_index, tensor_index)\u001b[0m\n\u001b[1;32m   1351\u001b[0m             \u001b[0mtensor_index\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mnode\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtensor_indices\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mi\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1352\u001b[0m             build_map(x, finished_nodes, nodes_in_progress, layer,\n\u001b[0;32m-> 1353\u001b[0;31m                       node_index, tensor_index)\n\u001b[0m\u001b[1;32m   1354\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1355\u001b[0m         \u001b[0mfinished_nodes\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0madd\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mnode\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/keras/engine/network.py\u001b[0m in \u001b[0;36mbuild_map\u001b[0;34m(tensor, finished_nodes, nodes_in_progress, layer, node_index, tensor_index)\u001b[0m\n\u001b[1;32m   1351\u001b[0m             \u001b[0mtensor_index\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mnode\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtensor_indices\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mi\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1352\u001b[0m             build_map(x, finished_nodes, nodes_in_progress, layer,\n\u001b[0;32m-> 1353\u001b[0;31m                       node_index, tensor_index)\n\u001b[0m\u001b[1;32m   1354\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1355\u001b[0m         \u001b[0mfinished_nodes\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0madd\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mnode\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/keras/engine/network.py\u001b[0m in \u001b[0;36mbuild_map\u001b[0;34m(tensor, finished_nodes, nodes_in_progress, layer, node_index, tensor_index)\u001b[0m\n\u001b[1;32m   1351\u001b[0m             \u001b[0mtensor_index\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mnode\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtensor_indices\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mi\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1352\u001b[0m             build_map(x, finished_nodes, nodes_in_progress, layer,\n\u001b[0;32m-> 1353\u001b[0;31m                       node_index, tensor_index)\n\u001b[0m\u001b[1;32m   1354\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1355\u001b[0m         \u001b[0mfinished_nodes\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0madd\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mnode\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/keras/engine/network.py\u001b[0m in \u001b[0;36mbuild_map\u001b[0;34m(tensor, finished_nodes, nodes_in_progress, layer, node_index, tensor_index)\u001b[0m\n\u001b[1;32m   1351\u001b[0m             \u001b[0mtensor_index\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mnode\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtensor_indices\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mi\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1352\u001b[0m             build_map(x, finished_nodes, nodes_in_progress, layer,\n\u001b[0;32m-> 1353\u001b[0;31m                       node_index, tensor_index)\n\u001b[0m\u001b[1;32m   1354\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1355\u001b[0m         \u001b[0mfinished_nodes\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0madd\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mnode\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/keras/engine/network.py\u001b[0m in \u001b[0;36mbuild_map\u001b[0;34m(tensor, finished_nodes, nodes_in_progress, layer, node_index, tensor_index)\u001b[0m\n\u001b[1;32m   1351\u001b[0m             \u001b[0mtensor_index\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mnode\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtensor_indices\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mi\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1352\u001b[0m             build_map(x, finished_nodes, nodes_in_progress, layer,\n\u001b[0;32m-> 1353\u001b[0;31m                       node_index, tensor_index)\n\u001b[0m\u001b[1;32m   1354\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1355\u001b[0m         \u001b[0mfinished_nodes\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0madd\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mnode\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/keras/engine/network.py\u001b[0m in \u001b[0;36mbuild_map\u001b[0;34m(tensor, finished_nodes, nodes_in_progress, layer, node_index, tensor_index)\u001b[0m\n\u001b[1;32m   1351\u001b[0m             \u001b[0mtensor_index\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mnode\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtensor_indices\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mi\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1352\u001b[0m             build_map(x, finished_nodes, nodes_in_progress, layer,\n\u001b[0;32m-> 1353\u001b[0;31m                       node_index, tensor_index)\n\u001b[0m\u001b[1;32m   1354\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1355\u001b[0m         \u001b[0mfinished_nodes\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0madd\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mnode\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/keras/engine/network.py\u001b[0m in \u001b[0;36mbuild_map\u001b[0;34m(tensor, finished_nodes, nodes_in_progress, layer, node_index, tensor_index)\u001b[0m\n\u001b[1;32m   1351\u001b[0m             \u001b[0mtensor_index\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mnode\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtensor_indices\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mi\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1352\u001b[0m             build_map(x, finished_nodes, nodes_in_progress, layer,\n\u001b[0;32m-> 1353\u001b[0;31m                       node_index, tensor_index)\n\u001b[0m\u001b[1;32m   1354\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1355\u001b[0m         \u001b[0mfinished_nodes\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0madd\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mnode\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/keras/engine/network.py\u001b[0m in \u001b[0;36mbuild_map\u001b[0;34m(tensor, finished_nodes, nodes_in_progress, layer, node_index, tensor_index)\u001b[0m\n\u001b[1;32m   1323\u001b[0m             \u001b[0mValueError\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0;32mif\u001b[0m \u001b[0ma\u001b[0m \u001b[0mcycle\u001b[0m \u001b[0;32mis\u001b[0m \u001b[0mdetected\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1324\u001b[0m         \"\"\"\n\u001b[0;32m-> 1325\u001b[0;31m         \u001b[0mnode\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mlayer\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_inbound_nodes\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mnode_index\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   1326\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1327\u001b[0m         \u001b[0;31m# Prevent cycles.\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;31mAttributeError\u001b[0m: 'NoneType' object has no attribute '_inbound_nodes'"
          ]
        }
      ]
    },
    {
      "metadata": {
        "id": "WE7b72OOFups",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "plot_model(modelFilt, to_file=\"FiltModel.png\", show_shapes=True, show_layer_names=True)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "_lY9fuSspeIA",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 850
        },
        "outputId": "945845f0-f067-43c0-d396-73e0c2bcc880"
      },
      "cell_type": "code",
      "source": [
        "# set weights again\n",
        "loaded_model_low_accuracy.summary()"
      ],
      "execution_count": 144,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "input_1 (InputLayer)         (None, 32, 32, 3)         0         \n",
            "_________________________________________________________________\n",
            "conv2d_1 (Conv2D)            (None, 32, 32, 32)        896       \n",
            "_________________________________________________________________\n",
            "batch_normalization_1 (Batch (None, 32, 32, 32)        128       \n",
            "_________________________________________________________________\n",
            "conv2d_2 (Conv2D)            (None, 32, 32, 32)        9248      \n",
            "_________________________________________________________________\n",
            "batch_normalization_2 (Batch (None, 32, 32, 32)        128       \n",
            "_________________________________________________________________\n",
            "max_pooling2d_1 (MaxPooling2 (None, 16, 16, 32)        0         \n",
            "_________________________________________________________________\n",
            "dropout_1 (Dropout)          (None, 16, 16, 32)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_3 (Conv2D)            (None, 16, 16, 64)        18496     \n",
            "_________________________________________________________________\n",
            "batch_normalization_3 (Batch (None, 16, 16, 64)        256       \n",
            "_________________________________________________________________\n",
            "conv2d_4 (Conv2D)            (None, 16, 16, 64)        36928     \n",
            "_________________________________________________________________\n",
            "batch_normalization_4 (Batch (None, 16, 16, 64)        256       \n",
            "_________________________________________________________________\n",
            "max_pooling2d_2 (MaxPooling2 (None, 8, 8, 64)          0         \n",
            "_________________________________________________________________\n",
            "dropout_2 (Dropout)          (None, 8, 8, 64)          0         \n",
            "_________________________________________________________________\n",
            "conv2d_5 (Conv2D)            (None, 8, 8, 128)         73856     \n",
            "_________________________________________________________________\n",
            "batch_normalization_5 (Batch (None, 8, 8, 128)         512       \n",
            "_________________________________________________________________\n",
            "conv2d_6 (Conv2D)            (None, 8, 8, 128)         147584    \n",
            "_________________________________________________________________\n",
            "batch_normalization_6 (Batch (None, 8, 8, 128)         512       \n",
            "_________________________________________________________________\n",
            "max_pooling2d_3 (MaxPooling2 (None, 4, 4, 128)         0         \n",
            "_________________________________________________________________\n",
            "dropout_3 (Dropout)          (None, 4, 4, 128)         0         \n",
            "_________________________________________________________________\n",
            "flatten_1 (Flatten)          (None, 2048)              0         \n",
            "_________________________________________________________________\n",
            "dense_1 (Dense)              (None, 3)                 6147      \n",
            "=================================================================\n",
            "Total params: 294,947\n",
            "Trainable params: 294,051\n",
            "Non-trainable params: 896\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "Bn5PhIlAqd8k",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "# loaded_model_low_accuracy.layers[1].get_weights()\n",
        "# loaded_model_low_accuracy.layers[0].get_weights()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "M48EuB2YpeEq",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "modelFilt.layers[1].set_weights(loaded_model_low_accuracy.layers[1].get_weights())\n",
        "modelFilt.layers[3].set_weights(loaded_model_low_accuracy.layers[2].get_weights())\n",
        "modelFilt.layers[5].set_weights(loaded_model_low_accuracy.layers[3].get_weights())\n",
        "\n",
        "\n",
        "modelFilt.layers[2].set_weights(loaded_model_high_accuracy.layers[1].get_weights())\n",
        "modelFilt.layers[4].set_weights(loaded_model_high_accuracy.layers[2].get_weights())\n",
        "modelFilt.layers[6].set_weights(loaded_model_high_accuracy.layers[3].get_weights())"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "-2wj7j0_v8uR",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "# modelFilt.layers[2].set_weights(loaded_model_low_accuracy.layers[1].get_weights())\n",
        "# modelFilt.layers[4].set_weights(loaded_model_low_accuracy.layers[2].get_weights())\n",
        "# modelFilt.layers[6].set_weights(loaded_model_low_accuracy.layers[3].get_weights())\n",
        "\n",
        "\n",
        "# modelFilt.layers[1].set_weights(loaded_model_low_accuracy.layers[1].get_weights())\n",
        "# modelFilt.layers[3].set_weights(loaded_model_low_accuracy.layers[2].get_weights())\n",
        "# modelFilt.layers[5].set_weights(loaded_model_low_accuracy.layers[3].get_weights())"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "NrTPnsVBtX05",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "# layer_dict_modelFilt['conv2d_1'].set_weights(loaded_model_low_accuracy.layers[1].get_weights())\n",
        "# layer_dict_modelFilt['batch_normalization_5'].set_weights(loaded_model_low_accuracy.layers[2].get_weights())\n",
        "# layer_dict_modelFilt['conv2d_2'].set_weights(loaded_model_low_accuracy.layers[3].get_weights())\n",
        "\n",
        "\n",
        "# layer_dict_modelFilt['HighAccuracyConv1'].set_weights(loaded_model_low_accuracy.layers[1].get_weights())\n",
        "# layer_dict_modelFilt['batch_normalization_3'].set_weights(loaded_model_low_accuracy.layers[2].get_weights())\n",
        "# layer_dict_modelFilt['HighAccuracyConv2'].set_weights(loaded_model_low_accuracy.layers[3].get_weights())"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "y3pdihEIpeB2",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "a0b4AZgQFum1",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "# # set weights of some of the layers and make it non trainable!\n",
        "# layer_dict_low = dict([(layer.name, layer) for layer in loaded_model_low_accuracy.layers])\n",
        "# layer_dict_modelFilt = dict([(layer.name, layer) for layer in modelFilt.layers])\n",
        "\n",
        "\n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "o-nabstBFukL",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "# weights_low_conv2d1 = layer_dict_low['conv2d_1'].get_weights()\n",
        "# layer_dict_modelFilt['conv2d_1'].set_weights(weights_low_conv2d1)\n",
        "\n",
        "# weights_low_bn1 = layer_dict_low['batch_normalization_1'].get_weights()\n",
        "# layer_dict_modelFilt['batch_normalization_5'].set_weights(weights_low_bn1)\n",
        "\n",
        "# weights_low_conv2d_2 = layer_dict_low['conv2d_2'].get_weights()\n",
        "# layer_dict_modelFilt['conv2d_2'].set_weights(weights_low_conv2d_2)\n",
        "\n",
        "# modelFilt.layers[1].set_weights(weights_low_conv2d1)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "7OoELuuikAn6",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "# modelFilt"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "wTCdHBwFZ4at",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "# set weights of some of the layers and make it non trainable!\n",
        "# layer_dict_high= dict([(layer.name, layer) for layer in loaded_model_high_accuracy.layers])\n",
        "# layer_dict_modelFilt = dict([(layer.name, layer) for layer in modelFilt.layers])"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "UXkQ68OyZnI1",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "# weights_highconv2d1 = layer_dict_high['conv2d_1'].get_weights()\n",
        "# layer_dict_modelFilt['HighAccuracyConv1'].set_weights(weights_highconv2d1)\n",
        "\n",
        "# weights_high_bn1 = layer_dict_high['batch_normalization_1'].get_weights()\n",
        "# layer_dict_modelFilt['batch_normalization_3'].set_weights(weights_high_bn1)\n",
        "\n",
        "# weights_high_conv2d_2 = layer_dict_high['conv2d_2'].get_weights()\n",
        "# layer_dict_modelFilt['HighAccuracyConv2'].set_weights(weights_low_conv2d_2)\n",
        "# ### need to modify and zero put non redundant weights"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "UevpvtNEo7Zz",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 326
        },
        "outputId": "b5bd4bbe-7402-4738-be45-29f1abd88b71"
      },
      "cell_type": "code",
      "source": [
        "# np.allclose(weights_highconv2d1,layer_dict_modelFilt['HighAccuracyConv1'].get_weights())"
      ],
      "execution_count": 37,
      "outputs": [
        {
          "output_type": "error",
          "ename": "TypeError",
          "evalue": "ignored",
          "traceback": [
            "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
            "\u001b[0;31mTypeError\u001b[0m                                 Traceback (most recent call last)",
            "\u001b[0;32m<ipython-input-37-a3c022cac015>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m \u001b[0mnp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mallclose\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mweights_highconv2d1\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mlayer_dict_modelFilt\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m'HighAccuracyConv1'\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mget_weights\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
            "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/numpy/core/numeric.py\u001b[0m in \u001b[0;36mallclose\u001b[0;34m(a, b, rtol, atol, equal_nan)\u001b[0m\n\u001b[1;32m   2421\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   2422\u001b[0m     \"\"\"\n\u001b[0;32m-> 2423\u001b[0;31m     \u001b[0mres\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mall\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0misclose\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0ma\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mb\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mrtol\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mrtol\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0matol\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0matol\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mequal_nan\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mequal_nan\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   2424\u001b[0m     \u001b[0;32mreturn\u001b[0m \u001b[0mbool\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mres\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   2425\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/numpy/core/numeric.py\u001b[0m in \u001b[0;36misclose\u001b[0;34m(a, b, rtol, atol, equal_nan)\u001b[0m\n\u001b[1;32m   2519\u001b[0m     \u001b[0my\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0marray\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0my\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mdtype\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mdt\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mcopy\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mFalse\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0msubok\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mTrue\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   2520\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 2521\u001b[0;31m     \u001b[0mxfin\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0misfinite\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mx\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   2522\u001b[0m     \u001b[0myfin\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0misfinite\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0my\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   2523\u001b[0m     \u001b[0;32mif\u001b[0m \u001b[0mall\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mxfin\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;32mand\u001b[0m \u001b[0mall\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0myfin\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;31mTypeError\u001b[0m: ufunc 'isfinite' not supported for the input types, and the inputs could not be safely coerced to any supported types according to the casting rule ''safe''"
          ]
        }
      ]
    },
    {
      "metadata": {
        "id": "Y781xnnuorld",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "# weights_highconv2d1"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "at5JqaEWojaV",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "# layer_dict_modelFilt['HighAccuracyConv1'].get_weights()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "wm5P4A9EY0th",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "# loaded_model_low_accuracy.summary()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "Lm7AFEZibLP9",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "# modelFilt.layers"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "jKdEeJCdzLNt",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 119
        },
        "outputId": "84a337ce-b4e2-4b7b-d6b8-5d4069b1c1da"
      },
      "cell_type": "code",
      "source": [
        "s = -1\n",
        "for layer in modelFilt.layers[0:7]:\n",
        "    s = s+1\n",
        "    if (s%2) !=0:\n",
        "      print(s)\n",
        "      print(layer)\n",
        "    layer.trainable = False"
      ],
      "execution_count": 150,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "1\n",
            "<keras.layers.convolutional.Conv2D object at 0x7f35b40904e0>\n",
            "3\n",
            "<keras.layers.normalization.BatchNormalization object at 0x7f35b40a57f0>\n",
            "5\n",
            "<keras.layers.convolutional.Conv2D object at 0x7f35b3f8aa58>\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "Uk2JwzAGbCj0",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 119
        },
        "outputId": "fbd74c22-ca4d-47d5-c860-92d024a3b53d"
      },
      "cell_type": "code",
      "source": [
        "s = -1\n",
        "for layer in modelFilt.layers[0:7]:\n",
        "    s = s+1\n",
        "    if (s%2) !=0:\n",
        "      print(s)\n",
        "      print(layer)\n",
        "      layer.trainable = False"
      ],
      "execution_count": 105,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "1\n",
            "<keras.layers.convolutional.Conv2D object at 0x7f35c14d33c8>\n",
            "3\n",
            "<keras.layers.normalization.BatchNormalization object at 0x7f35c14e5710>\n",
            "5\n",
            "<keras.layers.convolutional.Conv2D object at 0x7f35c13cbc18>\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "7EozUm-dSIx2",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1057
        },
        "outputId": "fa7fbb03-0436-4ccd-e5ac-6cea26f93b42"
      },
      "cell_type": "code",
      "source": [
        "modelFilt.summary()"
      ],
      "execution_count": 151,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "__________________________________________________________________________________________________\n",
            "Layer (type)                    Output Shape         Param #     Connected to                     \n",
            "==================================================================================================\n",
            "input_1 (InputLayer)            (None, 32, 32, 3)    0                                            \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_7 (Conv2D)               (None, 32, 32, 32)   896         input_1[0][0]                    \n",
            "__________________________________________________________________________________________________\n",
            "HighAccuracyConv1 (Conv2D)      (None, 32, 32, 32)   896         input_1[0][0]                    \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_15 (BatchNo (None, 32, 32, 32)   128         conv2d_7[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_13 (BatchNo (None, 32, 32, 32)   128         HighAccuracyConv1[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_8 (Conv2D)               (None, 32, 32, 32)   9248        batch_normalization_15[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "HighAccuracyConv2 (Conv2D)      (None, 32, 32, 32)   9248        batch_normalization_13[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "concatenate_2 (Concatenate)     (None, 32, 32, 64)   0           conv2d_8[0][0]                   \n",
            "                                                                 HighAccuracyConv2[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_16 (BatchNo (None, 32, 32, 64)   256         concatenate_2[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "max_pooling2d_4 (MaxPooling2D)  (None, 16, 16, 64)   0           batch_normalization_16[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "dropout_4 (Dropout)             (None, 16, 16, 64)   0           max_pooling2d_4[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_9 (Conv2D)               (None, 16, 16, 64)   36928       dropout_4[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_17 (BatchNo (None, 16, 16, 64)   256         conv2d_9[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_10 (Conv2D)              (None, 16, 16, 64)   36928       batch_normalization_17[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_18 (BatchNo (None, 16, 16, 64)   256         conv2d_10[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "max_pooling2d_5 (MaxPooling2D)  (None, 8, 8, 64)     0           batch_normalization_18[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "dropout_5 (Dropout)             (None, 8, 8, 64)     0           max_pooling2d_5[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_11 (Conv2D)              (None, 8, 8, 128)    73856       dropout_5[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_19 (BatchNo (None, 8, 8, 128)    512         conv2d_11[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2d_12 (Conv2D)              (None, 8, 8, 128)    147584      batch_normalization_19[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "batch_normalization_20 (BatchNo (None, 8, 8, 128)    512         conv2d_12[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "max_pooling2d_6 (MaxPooling2D)  (None, 4, 4, 128)    0           batch_normalization_20[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "dropout_6 (Dropout)             (None, 4, 4, 128)    0           max_pooling2d_6[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "flatten_2 (Flatten)             (None, 2048)         0           dropout_6[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "dense_2 (Dense)                 (None, 3)            6147        flatten_2[0][0]                  \n",
            "==================================================================================================\n",
            "Total params: 344,195\n",
            "Trainable params: 322,755\n",
            "Non-trainable params: 21,440\n",
            "__________________________________________________________________________________________________\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/keras/engine/training.py:490: UserWarning: Discrepancy between trainable weights and collected trainable weights, did you set `model.trainable` without calling `model.compile` after ?\n",
            "  'Discrepancy between trainable weights and collected trainable'\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "metadata": {
        "id": "fev8cENxbr-P",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "mcp_save = ModelCheckpoint('weights.hdf5', save_best_only=True, monitor='val_loss', mode='auto')"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "0BTKJa7Zbwtc",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "def lr_schedule(epoch):\n",
        "    lrate = 0.0001\n",
        "    if epoch > 75:\n",
        "        lrate = 0.0005\n",
        "    if epoch > 100:\n",
        "        lrate = 0.0003\n",
        "    return lrate"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "sIOMXQsjdSNX",
        "colab_type": "code",
        "outputId": "8c404fdf-e7c6-48a8-9112-7fb0966975a3",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 615
        }
      },
      "cell_type": "code",
      "source": [
        "modelFilt.fit_generator(datagen.flow(train_X, train_ground, batch_size=batch_size),\\\n",
        "                    steps_per_epoch=train_X.shape[0] // batch_size,epochs=30,\\\n",
        "                    verbose=1,validation_data=(valid_X,valid_ground),\n",
        "                    shuffle = True,callbacks=[mcp_save,LearningRateScheduler(lr_schedule)])"
      ],
      "execution_count": 153,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/keras/engine/training.py:490: UserWarning: Discrepancy between trainable weights and collected trainable weights, did you set `model.trainable` without calling `model.compile` after ?\n",
            "  'Discrepancy between trainable weights and collected trainable'\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/30\n",
            "187/187 [==============================] - 9s 47ms/step - loss: 1.0736 - acc: 0.6565 - val_loss: 1.8362 - val_acc: 0.6373\n",
            "Epoch 2/30\n",
            "187/187 [==============================] - 6s 33ms/step - loss: 0.6994 - acc: 0.7235 - val_loss: 2.0422 - val_acc: 0.6100\n",
            "Epoch 3/30\n",
            "187/187 [==============================] - 6s 34ms/step - loss: 0.6059 - acc: 0.7614 - val_loss: 0.6559 - val_acc: 0.7513\n",
            "Epoch 4/30\n",
            "187/187 [==============================] - 7s 38ms/step - loss: 0.5401 - acc: 0.7962 - val_loss: 7.1165 - val_acc: 0.3777\n",
            "Epoch 5/30\n",
            "187/187 [==============================] - 7s 37ms/step - loss: 0.5136 - acc: 0.8067 - val_loss: 4.8907 - val_acc: 0.4173\n",
            "Epoch 6/30\n",
            "187/187 [==============================] - 7s 38ms/step - loss: 0.4954 - acc: 0.8143 - val_loss: 10.3549 - val_acc: 0.3333\n",
            "Epoch 7/30\n",
            " 33/187 [====>.........................] - ETA: 4s - loss: 0.4724 - acc: 0.8329"
          ],
          "name": "stdout"
        },
        {
          "output_type": "error",
          "ename": "KeyboardInterrupt",
          "evalue": "ignored",
          "traceback": [
            "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
            "\u001b[0;31mKeyboardInterrupt\u001b[0m                         Traceback (most recent call last)",
            "\u001b[0;32m<ipython-input-153-3fcba61b69c7>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[1;32m      1\u001b[0m modelFilt.fit_generator(datagen.flow(train_X, train_ground, batch_size=batch_size),                    steps_per_epoch=train_X.shape[0] // batch_size,epochs=30,                    verbose=1,validation_data=(valid_X,valid_ground),\n\u001b[0;32m----> 2\u001b[0;31m                     shuffle = True,callbacks=[mcp_save,LearningRateScheduler(lr_schedule)])\n\u001b[0m",
            "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/keras/legacy/interfaces.py\u001b[0m in \u001b[0;36mwrapper\u001b[0;34m(*args, **kwargs)\u001b[0m\n\u001b[1;32m     89\u001b[0m                 warnings.warn('Update your `' + object_name + '` call to the ' +\n\u001b[1;32m     90\u001b[0m                               'Keras 2 API: ' + signature, stacklevel=2)\n\u001b[0;32m---> 91\u001b[0;31m             \u001b[0;32mreturn\u001b[0m \u001b[0mfunc\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     92\u001b[0m         \u001b[0mwrapper\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_original_function\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mfunc\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     93\u001b[0m         \u001b[0;32mreturn\u001b[0m \u001b[0mwrapper\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/keras/engine/training.py\u001b[0m in \u001b[0;36mfit_generator\u001b[0;34m(self, generator, steps_per_epoch, epochs, verbose, callbacks, validation_data, validation_steps, class_weight, max_queue_size, workers, use_multiprocessing, shuffle, initial_epoch)\u001b[0m\n\u001b[1;32m   1416\u001b[0m             \u001b[0muse_multiprocessing\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0muse_multiprocessing\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1417\u001b[0m             \u001b[0mshuffle\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mshuffle\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1418\u001b[0;31m             initial_epoch=initial_epoch)\n\u001b[0m\u001b[1;32m   1419\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1420\u001b[0m     \u001b[0;34m@\u001b[0m\u001b[0minterfaces\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlegacy_generator_methods_support\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/keras/engine/training_generator.py\u001b[0m in \u001b[0;36mfit_generator\u001b[0;34m(model, generator, steps_per_epoch, epochs, verbose, callbacks, validation_data, validation_steps, class_weight, max_queue_size, workers, use_multiprocessing, shuffle, initial_epoch)\u001b[0m\n\u001b[1;32m    179\u001b[0m             \u001b[0mbatch_index\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;36m0\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    180\u001b[0m             \u001b[0;32mwhile\u001b[0m \u001b[0msteps_done\u001b[0m \u001b[0;34m<\u001b[0m \u001b[0msteps_per_epoch\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 181\u001b[0;31m                 \u001b[0mgenerator_output\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mnext\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0moutput_generator\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    182\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    183\u001b[0m                 \u001b[0;32mif\u001b[0m \u001b[0;32mnot\u001b[0m \u001b[0mhasattr\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mgenerator_output\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m'__len__'\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/keras/utils/data_utils.py\u001b[0m in \u001b[0;36mget\u001b[0;34m(self)\u001b[0m\n\u001b[1;32m    683\u001b[0m         \u001b[0;32mtry\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    684\u001b[0m             \u001b[0;32mwhile\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mis_running\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 685\u001b[0;31m                 \u001b[0minputs\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mqueue\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mget\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mblock\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mTrue\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mget\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    686\u001b[0m                 \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mqueue\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtask_done\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    687\u001b[0m                 \u001b[0;32mif\u001b[0m \u001b[0minputs\u001b[0m \u001b[0;32mis\u001b[0m \u001b[0;32mnot\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/lib/python3.6/multiprocessing/pool.py\u001b[0m in \u001b[0;36mget\u001b[0;34m(self, timeout)\u001b[0m\n\u001b[1;32m    662\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    663\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0mget\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtimeout\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mNone\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 664\u001b[0;31m         \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mwait\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mtimeout\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    665\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0;32mnot\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mready\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    666\u001b[0m             \u001b[0;32mraise\u001b[0m \u001b[0mTimeoutError\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/lib/python3.6/multiprocessing/pool.py\u001b[0m in \u001b[0;36mwait\u001b[0;34m(self, timeout)\u001b[0m\n\u001b[1;32m    659\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    660\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0mwait\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtimeout\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mNone\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 661\u001b[0;31m         \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_event\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mwait\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mtimeout\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    662\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    663\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0mget\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtimeout\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mNone\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/lib/python3.6/threading.py\u001b[0m in \u001b[0;36mwait\u001b[0;34m(self, timeout)\u001b[0m\n\u001b[1;32m    549\u001b[0m             \u001b[0msignaled\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_flag\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    550\u001b[0m             \u001b[0;32mif\u001b[0m \u001b[0;32mnot\u001b[0m \u001b[0msignaled\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 551\u001b[0;31m                 \u001b[0msignaled\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_cond\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mwait\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mtimeout\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    552\u001b[0m             \u001b[0;32mreturn\u001b[0m \u001b[0msignaled\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    553\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/lib/python3.6/threading.py\u001b[0m in \u001b[0;36mwait\u001b[0;34m(self, timeout)\u001b[0m\n\u001b[1;32m    293\u001b[0m         \u001b[0;32mtry\u001b[0m\u001b[0;34m:\u001b[0m    \u001b[0;31m# restore state no matter what (e.g., KeyboardInterrupt)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    294\u001b[0m             \u001b[0;32mif\u001b[0m \u001b[0mtimeout\u001b[0m \u001b[0;32mis\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 295\u001b[0;31m                 \u001b[0mwaiter\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0macquire\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    296\u001b[0m                 \u001b[0mgotit\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;32mTrue\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    297\u001b[0m             \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;31mKeyboardInterrupt\u001b[0m: "
          ]
        }
      ]
    },
    {
      "metadata": {
        "id": "ie8_KBL8TIl6",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 346
        },
        "outputId": "e9dadd09-2000-4f72-d9e7-1974347cf229"
      },
      "cell_type": "code",
      "source": [
        "# now get DI score of a layer from high accuracy and low accuracy model\n",
        "layer_name = 'HighAccuracyConv2'\n",
        "discores_low_acc_model = createCNNLayerDIScorePlot(modelFilt, layer_name)"
      ],
      "execution_count": 154,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Mean of DI SCore:  0.47800322599191625\n",
            "STD of DI Score:  0.07607869969361732\n",
            "MAX of DI Score:  0.6407786286668374\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYUAAAEWCAYAAACJ0YulAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nQZLU+f+UYPbi8LQHCQAAAABJRU5ErkJggg==\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "metadata": {
        "id": "r2XDOWtOP2p7",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "modelFilt.summary()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "CXVt9e15sCf8",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 346
        },
        "outputId": "7e737ef7-5020-46e4-e233-d1caa2e064b1"
      },
      "cell_type": "code",
      "source": [
        "layer_name = 'conv2d_8'\n",
        "discores_low_acc_model = createCNNLayerDIScorePlot(modelFilt, layer_name)"
      ],
      "execution_count": 149,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Mean of DI SCore:  0.5925375772649611\n",
            "STD of DI Score:  0.056699249107349546\n",
            "MAX of DI Score:  0.6699787896299502\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYUAAAEWCAYAAACJ0YulAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nkqSOoSBJ6hgKkqSOoSBJ6vx/AQgiF0LFN4gAAAAASUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "metadata": {
        "id": "dxnMxP6cTGLO",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        " "
      ]
    },
    {
      "metadata": {
        "id": "_w7HlgF3XA9E",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "RX5m4EEvXA5g",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "72XqLffkXA20",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "EF-E1FlfXAgF",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        ""
      ]
    },
    {
      "metadata": {
        "id": "qyH9eKlYbXNn",
        "colab_type": "code",
        "outputId": "9629cc52-ce19-4bfb-ab0c-5d1c85e43c4a",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 2026
        }
      },
      "cell_type": "code",
      "source": [
        "modelFilt.fit_generator(datagen.flow(train_X, train_ground, batch_size=batch_size),\\\n",
        "                    steps_per_epoch=train_X.shape[0] // batch_size,epochs=60,\\\n",
        "                    verbose=1,validation_data=(valid_X,valid_ground),\n",
        "                    shuffle = True,callbacks=[mcp_save,LearningRateScheduler(lr_schedule)])\n",
        "#save to disk\n",
        "# model_json = model.to_json()\n",
        "# with open('modelHighAcc.json', 'w') as json_file:\n",
        "#     json_file.write(model_json)\n",
        "# model.save_weights('modelHighAcc.h5') \n",
        "# model.save('high_acc_complete_modelConv2d_2.h5')\n"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/60\n",
            "187/187 [==============================] - 10s 55ms/step - loss: 1.0914 - acc: 0.6170 - val_loss: 1.0533 - val_acc: 0.6503\n",
            "Epoch 2/60\n",
            "187/187 [==============================] - 6s 33ms/step - loss: 0.7965 - acc: 0.6631 - val_loss: 0.6973 - val_acc: 0.7247\n",
            "Epoch 3/60\n",
            "187/187 [==============================] - 6s 33ms/step - loss: 0.6907 - acc: 0.7025 - val_loss: 0.6277 - val_acc: 0.7327\n",
            "Epoch 4/60\n",
            "187/187 [==============================] - 6s 33ms/step - loss: 0.6372 - acc: 0.7331 - val_loss: 0.5845 - val_acc: 0.7633\n",
            "Epoch 5/60\n",
            "187/187 [==============================] - 6s 33ms/step - loss: 0.5893 - acc: 0.7575 - val_loss: 0.5723 - val_acc: 0.7650\n",
            "Epoch 6/60\n",
            "187/187 [==============================] - 6s 33ms/step - loss: 0.5638 - acc: 0.7772 - val_loss: 0.5688 - val_acc: 0.7923\n",
            "Epoch 7/60\n",
            "187/187 [==============================] - 7s 38ms/step - loss: 0.5404 - acc: 0.7868 - val_loss: 0.5032 - val_acc: 0.8123\n",
            "Epoch 8/60\n",
            "187/187 [==============================] - 7s 37ms/step - loss: 0.5651 - acc: 0.7914 - val_loss: 0.4749 - val_acc: 0.8183\n",
            "Epoch 9/60\n",
            "187/187 [==============================] - 6s 33ms/step - loss: 0.5130 - acc: 0.8015 - val_loss: 0.8505 - val_acc: 0.6970\n",
            "Epoch 10/60\n",
            "187/187 [==============================] - 7s 37ms/step - loss: 0.4912 - acc: 0.8123 - val_loss: 0.4684 - val_acc: 0.8337\n",
            "Epoch 11/60\n",
            "187/187 [==============================] - 7s 36ms/step - loss: 0.4880 - acc: 0.8147 - val_loss: 0.4659 - val_acc: 0.8307\n",
            "Epoch 12/60\n",
            "187/187 [==============================] - 6s 33ms/step - loss: 0.4772 - acc: 0.8246 - val_loss: 0.4656 - val_acc: 0.8333\n",
            "Epoch 13/60\n",
            "187/187 [==============================] - 6s 33ms/step - loss: 0.4620 - acc: 0.8308 - val_loss: 0.4341 - val_acc: 0.8480\n",
            "Epoch 14/60\n",
            "187/187 [==============================] - 6s 33ms/step - loss: 0.4480 - acc: 0.8366 - val_loss: 0.5995 - val_acc: 0.8087\n",
            "Epoch 15/60\n",
            "187/187 [==============================] - 6s 33ms/step - loss: 0.4506 - acc: 0.8375 - val_loss: 0.4449 - val_acc: 0.8377\n",
            "Epoch 16/60\n",
            "187/187 [==============================] - 6s 33ms/step - loss: 0.4393 - acc: 0.8422 - val_loss: 0.5139 - val_acc: 0.8280\n",
            "Epoch 17/60\n",
            "187/187 [==============================] - 6s 33ms/step - loss: 0.4462 - acc: 0.8391 - val_loss: 0.4592 - val_acc: 0.8350\n",
            "Epoch 18/60\n",
            "187/187 [==============================] - 6s 33ms/step - loss: 0.4315 - acc: 0.8464 - val_loss: 0.5755 - val_acc: 0.8043\n",
            "Epoch 19/60\n",
            "187/187 [==============================] - 6s 33ms/step - loss: 0.4305 - acc: 0.8475 - val_loss: 0.4241 - val_acc: 0.8500\n",
            "Epoch 20/60\n",
            "187/187 [==============================] - 6s 33ms/step - loss: 0.4261 - acc: 0.8501 - val_loss: 0.4542 - val_acc: 0.8373\n",
            "Epoch 21/60\n",
            "187/187 [==============================] - 6s 33ms/step - loss: 0.4137 - acc: 0.8554 - val_loss: 0.4838 - val_acc: 0.8317\n",
            "Epoch 22/60\n",
            "187/187 [==============================] - 6s 33ms/step - loss: 0.4216 - acc: 0.8521 - val_loss: 0.5241 - val_acc: 0.8177\n",
            "Epoch 23/60\n",
            "187/187 [==============================] - 7s 38ms/step - loss: 0.4239 - acc: 0.8544 - val_loss: 0.4090 - val_acc: 0.8617\n",
            "Epoch 24/60\n",
            "187/187 [==============================] - 7s 35ms/step - loss: 0.4111 - acc: 0.8630 - val_loss: 0.4724 - val_acc: 0.8287\n",
            "Epoch 25/60\n",
            "187/187 [==============================] - 6s 33ms/step - loss: 0.3932 - acc: 0.8648 - val_loss: 0.4638 - val_acc: 0.8450\n",
            "Epoch 26/60\n",
            "187/187 [==============================] - 6s 33ms/step - loss: 0.3906 - acc: 0.8677 - val_loss: 0.6440 - val_acc: 0.8327\n",
            "Epoch 27/60\n",
            "187/187 [==============================] - 6s 33ms/step - loss: 0.3948 - acc: 0.8681 - val_loss: 0.4341 - val_acc: 0.8533\n",
            "Epoch 28/60\n",
            "187/187 [==============================] - 6s 33ms/step - loss: 0.3927 - acc: 0.8655 - val_loss: 0.4833 - val_acc: 0.8410\n",
            "Epoch 29/60\n",
            "187/187 [==============================] - 6s 33ms/step - loss: 0.3919 - acc: 0.8715 - val_loss: 0.4792 - val_acc: 0.8440\n",
            "Epoch 30/60\n",
            "187/187 [==============================] - 6s 33ms/step - loss: 0.3878 - acc: 0.8689 - val_loss: 0.4615 - val_acc: 0.8557\n",
            "Epoch 31/60\n",
            "187/187 [==============================] - 6s 33ms/step - loss: 0.3759 - acc: 0.8753 - val_loss: 0.4198 - val_acc: 0.8637\n",
            "Epoch 32/60\n",
            "187/187 [==============================] - 7s 37ms/step - loss: 0.3850 - acc: 0.8729 - val_loss: 0.5197 - val_acc: 0.8247\n",
            "Epoch 33/60\n",
            "187/187 [==============================] - 6s 33ms/step - loss: 0.3701 - acc: 0.8810 - val_loss: 0.4236 - val_acc: 0.8650\n",
            "Epoch 34/60\n",
            "187/187 [==============================] - 6s 33ms/step - loss: 0.3829 - acc: 0.8744 - val_loss: 0.4045 - val_acc: 0.8663\n",
            "Epoch 35/60\n",
            "187/187 [==============================] - 7s 36ms/step - loss: 0.3761 - acc: 0.8809 - val_loss: 0.4072 - val_acc: 0.8637\n",
            "Epoch 36/60\n",
            "187/187 [==============================] - 7s 38ms/step - loss: 0.3773 - acc: 0.8762 - val_loss: 0.3791 - val_acc: 0.8743\n",
            "Epoch 37/60\n",
            "187/187 [==============================] - 6s 33ms/step - loss: 0.3643 - acc: 0.8834 - val_loss: 0.4467 - val_acc: 0.8547\n",
            "Epoch 38/60\n",
            "187/187 [==============================] - 6s 33ms/step - loss: 0.3615 - acc: 0.8849 - val_loss: 0.4095 - val_acc: 0.8637\n",
            "Epoch 39/60\n",
            "187/187 [==============================] - 6s 33ms/step - loss: 0.3662 - acc: 0.8799 - val_loss: 0.4111 - val_acc: 0.8677\n",
            "Epoch 40/60\n",
            "187/187 [==============================] - 6s 33ms/step - loss: 0.3657 - acc: 0.8847 - val_loss: 0.5137 - val_acc: 0.8497\n",
            "Epoch 41/60\n",
            "187/187 [==============================] - 6s 33ms/step - loss: 0.3730 - acc: 0.8814 - val_loss: 0.4308 - val_acc: 0.8660\n",
            "Epoch 42/60\n",
            "187/187 [==============================] - 6s 33ms/step - loss: 0.3685 - acc: 0.8839 - val_loss: 0.3999 - val_acc: 0.8767\n",
            "Epoch 43/60\n",
            "187/187 [==============================] - 6s 33ms/step - loss: 0.3533 - acc: 0.8926 - val_loss: 0.4592 - val_acc: 0.8573\n",
            "Epoch 44/60\n",
            "187/187 [==============================] - 6s 33ms/step - loss: 0.3510 - acc: 0.8917 - val_loss: 0.4648 - val_acc: 0.8547\n",
            "Epoch 45/60\n",
            "187/187 [==============================] - 6s 33ms/step - loss: 0.3586 - acc: 0.8880 - val_loss: 0.4407 - val_acc: 0.8600\n",
            "Epoch 46/60\n",
            "187/187 [==============================] - 6s 33ms/step - loss: 0.3548 - acc: 0.8914 - val_loss: 0.8102 - val_acc: 0.8277\n",
            "Epoch 47/60\n",
            "187/187 [==============================] - 6s 33ms/step - loss: 0.3541 - acc: 0.8898 - val_loss: 0.5405 - val_acc: 0.8200\n",
            "Epoch 48/60\n",
            " 98/187 [==============>...............] - ETA: 2s - loss: 0.3465 - acc: 0.8892"
          ],
          "name": "stdout"
        },
        {
          "output_type": "error",
          "ename": "KeyboardInterrupt",
          "evalue": "ignored",
          "traceback": [
            "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
            "\u001b[0;31mKeyboardInterrupt\u001b[0m                         Traceback (most recent call last)",
            "\u001b[0;32m<ipython-input-146-ec12c04c528b>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[1;32m      1\u001b[0m modelFilt.fit_generator(datagen.flow(train_X, train_ground, batch_size=batch_size),                    steps_per_epoch=train_X.shape[0] // batch_size,epochs=60,                    verbose=1,validation_data=(valid_X,valid_ground),\n\u001b[0;32m----> 2\u001b[0;31m                     shuffle = True,callbacks=[mcp_save,LearningRateScheduler(lr_schedule)])\n\u001b[0m\u001b[1;32m      3\u001b[0m \u001b[0;31m#save to disk\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      4\u001b[0m \u001b[0;31m# model_json = model.to_json()\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      5\u001b[0m \u001b[0;31m# with open('modelHighAcc.json', 'w') as json_file:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/keras/legacy/interfaces.py\u001b[0m in \u001b[0;36mwrapper\u001b[0;34m(*args, **kwargs)\u001b[0m\n\u001b[1;32m     89\u001b[0m                 warnings.warn('Update your `' + object_name + '` call to the ' +\n\u001b[1;32m     90\u001b[0m                               'Keras 2 API: ' + signature, stacklevel=2)\n\u001b[0;32m---> 91\u001b[0;31m             \u001b[0;32mreturn\u001b[0m \u001b[0mfunc\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     92\u001b[0m         \u001b[0mwrapper\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_original_function\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mfunc\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     93\u001b[0m         \u001b[0;32mreturn\u001b[0m \u001b[0mwrapper\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/keras/engine/training.py\u001b[0m in \u001b[0;36mfit_generator\u001b[0;34m(self, generator, steps_per_epoch, epochs, verbose, callbacks, validation_data, validation_steps, class_weight, max_queue_size, workers, use_multiprocessing, shuffle, initial_epoch)\u001b[0m\n\u001b[1;32m   1416\u001b[0m             \u001b[0muse_multiprocessing\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0muse_multiprocessing\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1417\u001b[0m             \u001b[0mshuffle\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mshuffle\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1418\u001b[0;31m             initial_epoch=initial_epoch)\n\u001b[0m\u001b[1;32m   1419\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1420\u001b[0m     \u001b[0;34m@\u001b[0m\u001b[0minterfaces\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlegacy_generator_methods_support\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/keras/engine/training_generator.py\u001b[0m in \u001b[0;36mfit_generator\u001b[0;34m(model, generator, steps_per_epoch, epochs, verbose, callbacks, validation_data, validation_steps, class_weight, max_queue_size, workers, use_multiprocessing, shuffle, initial_epoch)\u001b[0m\n\u001b[1;32m    179\u001b[0m             \u001b[0mbatch_index\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;36m0\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    180\u001b[0m             \u001b[0;32mwhile\u001b[0m \u001b[0msteps_done\u001b[0m \u001b[0;34m<\u001b[0m \u001b[0msteps_per_epoch\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 181\u001b[0;31m                 \u001b[0mgenerator_output\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mnext\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0moutput_generator\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    182\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    183\u001b[0m                 \u001b[0;32mif\u001b[0m \u001b[0;32mnot\u001b[0m \u001b[0mhasattr\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mgenerator_output\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m'__len__'\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/keras/utils/data_utils.py\u001b[0m in \u001b[0;36mget\u001b[0;34m(self)\u001b[0m\n\u001b[1;32m    683\u001b[0m         \u001b[0;32mtry\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    684\u001b[0m             \u001b[0;32mwhile\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mis_running\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 685\u001b[0;31m                 \u001b[0minputs\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mqueue\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mget\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mblock\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mTrue\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mget\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    686\u001b[0m                 \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mqueue\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtask_done\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    687\u001b[0m                 \u001b[0;32mif\u001b[0m \u001b[0minputs\u001b[0m \u001b[0;32mis\u001b[0m \u001b[0;32mnot\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/lib/python3.6/multiprocessing/pool.py\u001b[0m in \u001b[0;36mget\u001b[0;34m(self, timeout)\u001b[0m\n\u001b[1;32m    662\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    663\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0mget\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtimeout\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mNone\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 664\u001b[0;31m         \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mwait\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mtimeout\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    665\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0;32mnot\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mready\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    666\u001b[0m             \u001b[0;32mraise\u001b[0m \u001b[0mTimeoutError\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/lib/python3.6/multiprocessing/pool.py\u001b[0m in \u001b[0;36mwait\u001b[0;34m(self, timeout)\u001b[0m\n\u001b[1;32m    659\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    660\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0mwait\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtimeout\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mNone\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 661\u001b[0;31m         \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_event\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mwait\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mtimeout\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    662\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    663\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0mget\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtimeout\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mNone\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/lib/python3.6/threading.py\u001b[0m in \u001b[0;36mwait\u001b[0;34m(self, timeout)\u001b[0m\n\u001b[1;32m    549\u001b[0m             \u001b[0msignaled\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_flag\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    550\u001b[0m             \u001b[0;32mif\u001b[0m \u001b[0;32mnot\u001b[0m \u001b[0msignaled\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 551\u001b[0;31m                 \u001b[0msignaled\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_cond\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mwait\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mtimeout\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    552\u001b[0m             \u001b[0;32mreturn\u001b[0m \u001b[0msignaled\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    553\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/lib/python3.6/threading.py\u001b[0m in \u001b[0;36mwait\u001b[0;34m(self, timeout)\u001b[0m\n\u001b[1;32m    293\u001b[0m         \u001b[0;32mtry\u001b[0m\u001b[0;34m:\u001b[0m    \u001b[0;31m# restore state no matter what (e.g., KeyboardInterrupt)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    294\u001b[0m             \u001b[0;32mif\u001b[0m \u001b[0mtimeout\u001b[0m \u001b[0;32mis\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 295\u001b[0;31m                 \u001b[0mwaiter\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0macquire\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    296\u001b[0m                 \u001b[0mgotit\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;32mTrue\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    297\u001b[0m             \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;31mKeyboardInterrupt\u001b[0m: "
          ]
        }
      ]
    },
    {
      "metadata": {
        "id": "h7OTTp-uYZih",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "# loaded_model_low_accuracy.layers[1].get_weights()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "yDoD-D2pYLEd",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "# modelFilt.layers[1].get_weights()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "y-ssOHjPXTzE",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "# weights_low_conv2d1 = layer_dict_high['conv2d_1'].get_weights()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "KD0MAAeLFuhX",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "# loaded_model_low_accuracy.layers('conv2d_1').get_weights()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "A0ulN6H5FueV",
        "colab_type": "code",
        "outputId": "a289fa36-bbe6-4c59-e642-1c8d2d4b4a5a",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 374
        }
      },
      "cell_type": "code",
      "source": [
        "# weights_low_bn1 = layer_dict_high['batch_normalization_1'].get_weights()\n",
        "# modelFilt.layers[3].set_weights(weights_low_conv2d1)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "{'batch_normalization_1': <keras.layers.normalization.BatchNormalization at 0x7f1533511438>,\n",
              " 'batch_normalization_2': <keras.layers.normalization.BatchNormalization at 0x7f15335117f0>,\n",
              " 'batch_normalization_3': <keras.layers.normalization.BatchNormalization at 0x7f1533511ba8>,\n",
              " 'batch_normalization_4': <keras.layers.normalization.BatchNormalization at 0x7f1533511e80>,\n",
              " 'batch_normalization_5': <keras.layers.normalization.BatchNormalization at 0x7f153352d278>,\n",
              " 'batch_normalization_6': <keras.layers.normalization.BatchNormalization at 0x7f153352d550>,\n",
              " 'conv2d_1': <keras.layers.convolutional.Conv2D at 0x7f153350c0f0>,\n",
              " 'conv2d_2': <keras.layers.convolutional.Conv2D at 0x7f1533511668>,\n",
              " 'conv2d_3': <keras.layers.convolutional.Conv2D at 0x7f15335119e8>,\n",
              " 'conv2d_4': <keras.layers.convolutional.Conv2D at 0x7f1533511cc0>,\n",
              " 'conv2d_5': <keras.layers.convolutional.Conv2D at 0x7f153352d0b8>,\n",
              " 'conv2d_6': <keras.layers.convolutional.Conv2D at 0x7f153352d390>,\n",
              " 'dense_1': <keras.layers.core.Dense at 0x7f153352d7b8>,\n",
              " 'dropout_1': <keras.layers.core.Dropout at 0x7f15335119b0>,\n",
              " 'dropout_2': <keras.layers.core.Dropout at 0x7f153352d080>,\n",
              " 'dropout_3': <keras.layers.core.Dropout at 0x7f153352d710>,\n",
              " 'flatten_1': <keras.layers.core.Flatten at 0x7f153352d748>,\n",
              " 'input_1': <keras.engine.input_layer.InputLayer at 0x7f153350c438>,\n",
              " 'max_pooling2d_1': <keras.layers.pooling.MaxPooling2D at 0x7f1533511908>,\n",
              " 'max_pooling2d_2': <keras.layers.pooling.MaxPooling2D at 0x7f1533511f98>,\n",
              " 'max_pooling2d_3': <keras.layers.pooling.MaxPooling2D at 0x7f153352d668>}"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 111
        }
      ]
    },
    {
      "metadata": {
        "id": "TEkg52f_HxEg",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "## Ignore below"
      ]
    },
    {
      "metadata": {
        "id": "_iOvwS5SJVT8",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "def cifar10ClassifierTransfer(input_img, highAccInput):\n",
        "  \n",
        "  # Add additional inputs to \n",
        "  \n",
        "  conv1_1 = (Conv2D(32, (3,3), padding='same', kernel_regularizer=regularizers.l2(weight_decay), input_shape=x_train.shape[1:], activation='elu'))(input_img)\n",
        "  conv1_2 = BatchNormalization()(conv1_1)\n",
        "  conv2_1 = (Conv2D(32, (3,3), padding='same', activation= 'elu',kernel_regularizer=regularizers.l2(weight_decay)))(conv1_2)\n",
        "  # now concat secondary input that corresponds to high accuracy n/w to this\n",
        "    \n",
        "  concated_layer = keras.layers.merge.Concatenate()([conv2_1, highAccInput])\n",
        "  conv2_2 = BatchNormalization()(concated_layer)\n",
        "  pool2_3 = MaxPooling2D(pool_size=(2, 2))(conv2_2) #14 x 14 x 32\n",
        "  drop2_4= Dropout(.2)(pool2_3)\n",
        "\n",
        "  conv3_1 = (Conv2D(64, (3,3), padding='same', kernel_regularizer=regularizers.l2(weight_decay), activation='elu'))(drop2_4)\n",
        "  conv3_2 = BatchNormalization()(conv3_1)\n",
        "  conv4_1 = (Conv2D(64, (3,3), padding='same', activation= 'elu',kernel_regularizer=regularizers.l2(weight_decay)))(conv3_2)\n",
        "  conv4_2 = BatchNormalization()(conv4_1)\n",
        "  pool4_3 = MaxPooling2D(pool_size=(2, 2))(conv4_2) #14 x 14 x 32\n",
        "  drop4_4= Dropout(.3)(pool4_3)\n",
        "\n",
        "  conv5_1 = (Conv2D(128, (3,3), padding='same', kernel_regularizer=regularizers.l2(weight_decay), activation='elu'))(drop4_4)\n",
        "  conv5_2 = BatchNormalization()(conv5_1)\n",
        "  conv6_1 = (Conv2D(128, (3,3), padding='same', activation= 'elu',kernel_regularizer=regularizers.l2(weight_decay)))(conv5_2)\n",
        "  conv6_2 = BatchNormalization()(conv6_1)\n",
        "  pool6_3 = MaxPooling2D(pool_size=(2, 2))(conv6_2) #14 x 14 x 32\n",
        "  drop6_4= Dropout(.4)(pool6_3)\n",
        "\n",
        "  flat = Flatten()(drop6_4)\n",
        "  out = Dense(num_classes, activation='softmax')(flat)\n",
        "  \n",
        "  return out"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "PyCdAWfOBhSi",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "def featureExtractor(input_img, layer_name):\n",
        "  get_cnn1_output_function = K.function([trained_model.layers[0].input],\n",
        "                                  [trained_model.get_layer(layer_name).output])   #get_layer test_three_label\n",
        "  cnn1_output = get_cnn1_output_function([train_three_data])[0]\n",
        "  return cnn1_output"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "lYM0VB8wBhP8",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "def cifar10ClassifierTransfer(input_img, conv2_outputs):\n",
        "  \n",
        "  # Add additional inputs to \n",
        "  \n",
        "  conv1_1 = (Conv2D(32, (3,3), padding='same', kernel_regularizer=regularizers.l2(weight_decay), input_shape=x_train.shape[1:], activation='elu'))(input_img)\n",
        "  conv1_2 = BatchNormalization()(conv1_1)\n",
        "  conv2_1 = (Conv2D(32, (3,3), padding='same', activation= 'elu',kernel_regularizer=regularizers.l2(weight_decay)))(conv1_2)\n",
        "  conv2_2 = BatchNormalization()(conv2_1)\n",
        "  pool2_3 = MaxPooling2D(pool_size=(2, 2))(conv2_2) #14 x 14 x 32\n",
        "  drop2_4= Dropout(.2)(pool2_3)\n",
        "\n",
        "  conv3_1 = (Conv2D(64, (3,3), padding='same', kernel_regularizer=regularizers.l2(weight_decay), activation='elu'))(drop2_4)\n",
        "  conv3_2 = BatchNormalization()(conv3_1)\n",
        "  conv4_1 = (Conv2D(64, (3,3), padding='same', activation= 'elu',kernel_regularizer=regularizers.l2(weight_decay)))(conv3_2)\n",
        "  conv4_2 = BatchNormalization()(conv4_1)\n",
        "  pool4_3 = MaxPooling2D(pool_size=(2, 2))(conv4_2) #14 x 14 x 32\n",
        "  drop4_4= Dropout(.3)(pool4_3)\n",
        "\n",
        "  conv5_1 = (Conv2D(128, (3,3), padding='same', kernel_regularizer=regularizers.l2(weight_decay), activation='elu'))(drop4_4)\n",
        "  conv5_2 = BatchNormalization()(conv5_1)\n",
        "  conv6_1 = (Conv2D(128, (3,3), padding='same', activation= 'elu',kernel_regularizer=regularizers.l2(weight_decay)))(conv5_2)\n",
        "  conv6_2 = BatchNormalization()(conv6_1)\n",
        "  pool6_3 = MaxPooling2D(pool_size=(2, 2))(conv6_2) #14 x 14 x 32\n",
        "  drop6_4= Dropout(.4)(pool6_3)\n",
        "\n",
        "  flat = Flatten()(drop6_4)\n",
        "  out = Dense(num_classes, activation='softmax')(flat)\n",
        "  \n",
        "  return out"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "sXXygEe9BhNK",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "LDTPGc0ABhKO",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "gsrj25pSoRh_",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "def highAccuracyModelTillConv2(input_img, channel_indices_first_layer, channel_indices_second_layer): \n",
        "  conv1_1 = (Conv2D(32, (3,3), padding='same', kernel_regularizer=regularizers.l2(weight_decay), input_shape=x_train.shape[1:], activation='elu', name = 'HighAccuracyConv1'))(input_img)\n",
        "  conv1_2 = BatchNormalization()(conv1_1)\n",
        "  conv2_1 = (Conv2D(32, (3,3), padding='same', activation= 'elu', name = 'HighAccuracyConv2',kernel_regularizer=regularizers.l2(weight_decay)))(conv1_2)\n",
        "  conv2_2 = BatchNormalization()(conv2_1)\n",
        "#   pool1 = MaxPooling2D(pool_size=(2, 2))(conv1) #14 x 14 x 32\n",
        "#   drop1= Dropout(.2)(pool1)\n",
        "\n",
        "#   conv2 = (Conv2D(64, (3,3), padding='same', kernel_regularizer=regularizers.l2(weight_decay), activation='elu'))(drop1)\n",
        "#   conv2 = BatchNormalization()(conv2)\n",
        "#   conv2 = (Conv2D(64, (3,3), padding='same', activation= 'elu',kernel_regularizer=regularizers.l2(weight_decay)))(conv2)\n",
        "#   conv2 = BatchNormalization()(conv2)\n",
        "#   pool2 = MaxPooling2D(pool_size=(2, 2))(conv2) #14 x 14 x 32\n",
        "#   drop2= Dropout(.3)(pool2)\n",
        "\n",
        "#   conv3 = (Conv2D(128, (3,3), padding='same', kernel_regularizer=regularizers.l2(weight_decay), activation='elu'))(drop2)\n",
        "#   conv3 = BatchNormalization()(conv3)\n",
        "#   conv3 = (Conv2D(128, (3,3), padding='same', activation= 'elu',kernel_regularizer=regularizers.l2(weight_decay)))(conv3)\n",
        "#   conv3 = BatchNormalization()(conv3)\n",
        "#   pool3 = MaxPooling2D(pool_size=(2, 2))(conv3) #14 x 14 x 32\n",
        "#   drop3= Dropout(.4)(pool3)\n",
        "\n",
        "#   flat = Flatten()(drop3)\n",
        "#   out = Dense(num_classes, activation='softmax')(flat)\n",
        "  \n",
        "  return conv1_1[:,:,:,channel_indices_first_layer], conv2_1[:,:,:,channel_indices_second_layer]"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "RsXlJYX62MB-",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    }
  ]
}