Introduzione_Keras_ITA.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "provenance": [],
      "collapsed_sections": [
        "tvmEIPc99ZM4",
        "yQPsHM4WNYje",
        "Vs3sxPYBC7zG",
        "ilNqwkVdFMxI",
        "7yU-TXQpHLFe",
        "YiN4QylsWCgt",
        "NZzbsnY6l156",
        "P5kgRQdhF5fw",
        "kEqxcoRtnp8B",
        "aqORmpYEHm3z",
        "q76XQ7ED5u9_",
        "sPjwGZdd5NsU",
        "2FOuNPqHNDGw",
        "vnZLq5sxNVal",
        "90Ciuc_oGoPC"
      ]
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "source": [
        "#Keras\n",
        "\n",
        "Mario Fiorino\n",
        "\n",
        "\n",
        "Le API di TensorFlow sono organizzate gerarchicamente, con quelle di alto livello che si basano su quelle di basso livello.\n",
        "\n",
        "I ricercatori di machine learning utilizzano le API di basso livello per creare ed esplorare nuovi algoritmi.\n",
        "\n",
        "tf.keras è un'API di alto livello che consente di costruire ed addestrare semplici modelli di machine learning in modo facile (User-Friendly ) e veloce.\n",
        "\n",
        "Ref:\n",
        "\n",
        "https://keras.io/guides/\n",
        "\n",
        "https://keras.io/getting_started/faq/\n",
        "\n",
        "https://github.com/keras-team/keras/tree/v3.1.1/keras\n",
        "\n",
        "\n",
        "\n",
        "\n",
        "*\n",
        "\n",
        "https://elearning.di.unipi.it/pluginfile.php/16483/course/section/1837/%5BISPR%5D%20Intro%20to%20Machine%20Learning%20with%20Keras.pdf\n",
        "\n",
        "\n",
        "*\n",
        "\n",
        "https://github.com/jeffheaton/t81_558_deep_learning\n",
        "\n",
        "\n",
        "\n",
        "https://developers.google.com/machine-learning/crash-course/first-steps-with-tensorflow/programming-exercises?hl=it\n",
        "\n",
        "\n",
        "\n",
        "*\n",
        "\n",
        "\n",
        "https://keras.io/guides/custom_train_step_in_tensorflow/\n",
        "\n",
        "\n",
        "https://keras.io/api/utils/python_utils/#sequence-class\n",
        "\n",
        "\n",
        "\n"
      ],
      "metadata": {
        "id": "vZnm-AjA3Q8Z"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "<div>\n",
        "<img src=\"data:image/png;base64,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\n",
        "\", width=\"500\">\n",
        "</div>"
      ],
      "metadata": {
        "id": "bz1Mf70r3_z1"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "Riproponiamo la stessa architettura di rete neurale vista nel notebook precedente (Introduzione TensorFlow_v2_parte2_ITA), avvalendoci però della libreria Keras per la sua implementazione.\n",
        "\n",
        "Per considerazioni teoriche si consiglia di consultare il notebook:\n",
        "\n",
        "Introduzione TensorFlow_v2_parte2_ITA"
      ],
      "metadata": {
        "id": "7nsIK2yi4B2Y"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "import tensorflow as tf\n",
        "print(\"TensorFlow version:\", tf.__version__)\n",
        "\n",
        "from keras import initializers\n",
        "import os\n",
        "import numpy as np\n",
        "import pandas as pd\n",
        "from matplotlib import pyplot as plt"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "49819075-834c-41a5-b11c-25bf1521336e",
        "id": "yy_eQoEBsHzp"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "TensorFlow version: 2.15.0\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "#### Chiarimenti tecnici su Keras"
      ],
      "metadata": {
        "id": "P5kgRQdhF5fw"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "# NN ad uno strato, 3 nodi, 2 input, funzione di attivazione : Sigmoid\n",
        "\n",
        "model = tf.keras.models.Sequential()\n",
        "\n",
        "model.add(tf.keras.layers.Dense(units=3,\n",
        "                                  input_dim=2,\n",
        "                                  kernel_initializer=initializers.RandomNormal(mean=0.0, stddev=0.1, seed=None),\n",
        "                                  bias_initializer=initializers.Zeros(),\n",
        "                                  activation='sigmoid') )\n",
        "\n",
        "y =model.predict([[0,2]])\n",
        "print(y)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "FqWQnmZo8Vnc",
        "outputId": "7fee61d9-5419-473c-ce88-5a7e8ecc405d"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "1/1 [==============================] - 0s 61ms/step\n",
            "[[0.4896491 0.5199425 0.5379917]]\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Come funzione Accuracy in Keras\n",
        "\n",
        "#https://keras.io/api/metrics/accuracy_metrics/#accuracy-class\n",
        "\n",
        "prev =     [[1], [1], [1], [4]]\n",
        "test_set = [[0], [2], [3], [4]]\n",
        "\n",
        "m = tf.keras.metrics.Accuracy()\n",
        "m.update_state(test_set, prev)\n",
        "m.result().numpy()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "8JMPKsvm84oJ",
        "outputId": "de1c29b0-c36c-4053-ad93-fc50763d650b"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0.25"
            ]
          },
          "metadata": {},
          "execution_count": 23
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "#Loss function e batch size in Keras\n",
        "#1.\n",
        "\n",
        "# https://keras.io/api/losses/\n",
        "\n",
        "\n",
        "k1 = tf.ones((1,))*5\n",
        "print(k1)\n",
        "k2 = tf.ones((1, ))*2\n",
        "print(k2)\n",
        "#z = tf.zeros((2,2 ))\n",
        "d = tf.keras.losses.MeanSquaredError(reduction='none')(k1, k2)\n",
        "# \"none\" means the loss instance will return the full array of sample losses.\n",
        "print(d,\"\\n\")  # (5-2)^2 -> 9/1 = 9\n",
        "\n",
        "\n",
        "k1 = tf.constant([0, 7, 5])\n",
        "print(k1)\n",
        "k2 = tf.constant([1, 2, 1])\n",
        "print(k2)\n",
        "d = tf.keras.losses.MeanSquaredError(reduction='none')(k1, k2)\n",
        "print(d,\"\\n\") # 1^2 + 5^2 + 4^2 = 42 -> 42/3 = 14\n",
        "\n",
        "\n",
        "k1 = tf.constant([[0, 5, 0],[1, 2, 3]])\n",
        "print(k1)\n",
        "k2 = tf.constant([[1, 1, 0],[0, 3, 0]])\n",
        "print(k2)\n",
        "d = tf.keras.losses.MeanSquaredError(reduction='none')(k1, k2)\n",
        "print(d,\"\\n\") # 1^2 + 4^2 + 0 ->  15/3 ; 1^2 + 1^2 + 3^2 -> 11/3"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Oo2v6QPLQdZH",
        "outputId": "580bd90d-af26-44a7-d258-4536774c9d33"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "tf.Tensor([5.], shape=(1,), dtype=float32)\n",
            "tf.Tensor([2.], shape=(1,), dtype=float32)\n",
            "tf.Tensor(9.0, shape=(), dtype=float32) \n",
            "\n",
            "tf.Tensor([0 7 5], shape=(3,), dtype=int32)\n",
            "tf.Tensor([1 2 1], shape=(3,), dtype=int32)\n",
            "tf.Tensor(14, shape=(), dtype=int32) \n",
            "\n",
            "tf.Tensor(\n",
            "[[0 5 0]\n",
            " [1 2 3]], shape=(2, 3), dtype=int32)\n",
            "tf.Tensor(\n",
            "[[1 1 0]\n",
            " [0 3 0]], shape=(2, 3), dtype=int32)\n",
            "tf.Tensor([5 3], shape=(2,), dtype=int32) \n",
            "\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "#Loss function e batch size in Keras\n",
        "#2.\n",
        "# \"sum_over_batch_size\" means the loss instance will return the average of the sample losses in the batch.\n",
        "\n",
        "k1 = tf.ones((1,))*5\n",
        "print(k1)\n",
        "k2 = tf.ones((1, ))*2\n",
        "print(k2)\n",
        "#z = tf.zeros((2,2 ))\n",
        "d = tf.keras.losses.MeanSquaredError(reduction='sum_over_batch_size')(k1, k2)\n",
        "# \"none\" means the loss instance will return the full array of per-sample losses.\n",
        "print(d,\"\\n\")  # (5-2)^2 -> 9/1 = 9\n",
        "\n",
        "\n",
        "k1 = tf.constant([0, 7, 5])\n",
        "print(k1)\n",
        "k2 = tf.constant([1, 2, 1])\n",
        "print(k2)\n",
        "d = tf.keras.losses.MeanSquaredError(reduction='sum_over_batch_size')(k1, k2)\n",
        "print(d,\"\\n\") # 1^2 + 5^2 + 4^2 = 42 -> 42/3 = 14\n",
        "\n",
        "\n",
        "k1 = tf.constant([[0, 5, 0],[1, 2, 3]])\n",
        "print(k1)\n",
        "k2 = tf.constant([[1, 1, 0],[0, 3, 0]])\n",
        "print(k2)\n",
        "d = tf.keras.losses.MeanSquaredError(reduction='sum_over_batch_size')(k1, k2)\n",
        "print(d,\"\\n\") # tf.Tensor([5 3] -> (5+3)/2\n",
        "\n",
        "\n",
        "#\n",
        "# NOTA IMPORTANTE\n",
        "# When using fit(), this difference is irrelevant since reduction is handled by the framework.\n",
        "#\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "QSOttDpee0ZL",
        "outputId": "07480aaa-05f6-4a43-d9e9-8bb3d319fcb3"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "tf.Tensor([5.], shape=(1,), dtype=float32)\n",
            "tf.Tensor([2.], shape=(1,), dtype=float32)\n",
            "tf.Tensor(9.0, shape=(), dtype=float32) \n",
            "\n",
            "tf.Tensor([0 7 5], shape=(3,), dtype=int32)\n",
            "tf.Tensor([1 2 1], shape=(3,), dtype=int32)\n",
            "tf.Tensor(14, shape=(), dtype=int32) \n",
            "\n",
            "tf.Tensor(\n",
            "[[0 5 0]\n",
            " [1 2 3]], shape=(2, 3), dtype=int32)\n",
            "tf.Tensor(\n",
            "[[1 1 0]\n",
            " [0 3 0]], shape=(2, 3), dtype=int32)\n",
            "tf.Tensor(4, shape=(), dtype=int32) \n",
            "\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "#Loss function e batch size in Keras\n",
        "#3.\n",
        "\n",
        "k1 = tf.ones((2,3))*5\n",
        "print(k1)\n",
        "k2 = tf.ones((2,3))*2\n",
        "print(k2)\n",
        "d = tf.keras.losses.MeanSquaredError(reduction='auto')(k1, k2)\n",
        "d1 = tf.keras.losses.MeanSquaredError()(k1, k2)\n",
        "print(d,\"\\n\",d1)\n",
        "\n",
        "# Default value is auto: reduction='auto'\n",
        "# auto indicates that the reduction option will be determined by the usage context.\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "ui6c7w-0ez20",
        "outputId": "16fe9f84-b14c-4bb1-b651-290664b052a4"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "tf.Tensor(\n",
            "[[5. 5. 5.]\n",
            " [5. 5. 5.]], shape=(2, 3), dtype=float32)\n",
            "tf.Tensor(\n",
            "[[2. 2. 2.]\n",
            " [2. 2. 2.]], shape=(2, 3), dtype=float32)\n",
            "tf.Tensor(9.0, shape=(), dtype=float32) \n",
            " tf.Tensor(9.0, shape=(), dtype=float32)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "####Importa il dataset MNIST ed esegui la Data Preparation"
      ],
      "metadata": {
        "id": "kEqxcoRtnp8B"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "(xtrain,label_train),(xtest,label_test)=tf.keras.datasets.mnist.load_data()\n",
        "\n",
        "num_pixels = (xtrain[0].shape[0]) * (xtrain[0].shape[0])\n",
        "print(num_pixels,\"\\n\")\n",
        "\n",
        "# flatten 28*28 pixel images to a 784 vector\n",
        "xtrain = xtrain.reshape(xtrain.shape[0], num_pixels).astype('float32')\n",
        "xtest = xtest.reshape(xtest.shape[0], num_pixels).astype('float32')\n",
        "\n",
        "# normalize inputs from 0-255 to 0-1\n",
        "xtrain = xtrain / 255\n",
        "xtest = xtest / 255\n",
        "print(xtrain[0].shape,\"\\n\")\n",
        "\n",
        "# one hot encode outputs\n",
        "label_train = tf.keras.utils.to_categorical(label_train) # utils is tools for creating one-hot encoding\n",
        "label_test = tf.keras.utils.to_categorical(label_test)\n",
        "print(label_train[0])"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "f0qCjTPvndBT",
        "outputId": "d410845b-15f1-46d3-9e4b-b0d99a93abdf"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/mnist.npz\n",
            "11490434/11490434 [==============================] - 1s 0us/step\n",
            "784 \n",
            "\n",
            "(784,) \n",
            "\n",
            "[0. 0. 0. 0. 0. 1. 0. 0. 0. 0.]\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "#### Definizioni dei parametri del processo di training"
      ],
      "metadata": {
        "id": "aqORmpYEHm3z"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "#Parametri\n",
        "\n",
        "#number of neurons in input layer\n",
        "num_input = 784\n",
        "\n",
        "#number of neurons in hidden layer 1\n",
        "num_hidden1 = 128\n",
        "\n",
        "#number of neurons in hidden layer 2\n",
        "num_hidden2 = 64\n",
        "\n",
        "#number of neurons in output layer\n",
        "num_output = 10\n",
        "\n",
        "flag_summery = True\n",
        "\n",
        "n_layers = [num_input, num_hidden1, num_hidden2, num_output]\n",
        "\n",
        "# Number of epochs\n",
        "num_iterations = 20\n",
        "\n",
        "learning_rate=0.001\n",
        "\n",
        "# Number of samples per gradient update\n",
        "my_batch_size= 16\n",
        "\n",
        "#\n",
        "#Suggerimenti da Google:\n",
        "#\n",
        "\n",
        "#Setting the batch size to a very small batch number can also cause instability.\n",
        "#First, try large batch size values. Then, decrease the batch size until you see degradation\n",
        "\n",
        "#If the training loss decreases too slowly, increase the learning rate. Note that setting the learning rate too high may also prevent training loss from converging.\n",
        "\n",
        "#If the training loss varies wildly (that is, the training loss jumps around), decrease the learning rate.\n",
        "\n",
        "#Lowering the learning rate while increasing the number of epochs or the batch size is often a good combination."
      ],
      "metadata": {
        "id": "zSgWtSvj396Z"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "#### Definizione dell'architettura della rete neurale"
      ],
      "metadata": {
        "id": "q76XQ7ED5u9_"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "#I modelli tf.keras più semplici sono Sequential() - Di fatto una pila lineare a più livelli.\n",
        "  # Un modello sequenziale può contenere uno o più layer.\n",
        "  # Per architetture più complesse, si può utilizzare: https://keras.io/guides/functional_api/\n",
        "model = tf.keras.models.Sequential()\n",
        "\n",
        "##Layer1\n",
        "# Dense layers, oppure, fully-connected layers :\n",
        "# Dense implementa l'operazione : output = function_activation(dot(input, kernel) + bias)\n",
        "# dove: kernel è la  weights matrix\n",
        "model.add(tf.keras.layers.Dense(units=n_layers[1],       # units = il numero dei nodi - ovvero l'output space -\n",
        "                                  input_dim=n_layers[0],   # input_dim = dimensioni del Input Layer, oppure, Number of Input Feature model\n",
        "                                  kernel_initializer=initializers.RandomNormal(mean=0.0, stddev=0.1, seed=None),\n",
        "                                  bias_initializer=initializers.Zeros(),\n",
        "                                  activation='relu') )\n",
        "\n",
        "\n",
        "##Layer2\n",
        "model.add(tf.keras.layers.Dense(units=n_layers[2],\n",
        "                                  input_dim=n_layers[1],\n",
        "                                  kernel_initializer=initializers.RandomNormal(mean=0.0, stddev=0.1, seed=None),\n",
        "                                  bias_initializer=initializers.Zeros(),\n",
        "                                  activation='relu') )\n",
        "\n",
        "\n",
        "##Layer3\n",
        "model.add(tf.keras.layers.Dense(units=n_layers[3],\n",
        "                                  input_dim=n_layers[2],\n",
        "                                  kernel_initializer=initializers.RandomNormal(mean=0.0, stddev=0.1, seed=None),\n",
        "                                  bias_initializer=initializers.Zeros(),\n",
        "                                  activation='softmax') )\n",
        "\n",
        "# Compile the model\n",
        "  # Ovvero :  configura il suo processo di apprendimento\n",
        "  # Configurare il processo di learning per ridurre al minimo la funzione di loss scelta.\n",
        "model.compile(optimizer=tf.keras.optimizers.Adam(learning_rate=learning_rate),\n",
        "                loss= tf.keras.losses.CategoricalCrossentropy(from_logits=False),\n",
        "                metrics=[tf.keras.metrics.CategoricalAccuracy()])\n",
        "\n",
        "\n",
        "  # Per stampa le info del modello creato\n",
        "if flag_summery == True:\n",
        "      model.summary()\n",
        "      #tf.keras.utils.plot_model(model)\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "WdbwFZkqqn9_",
        "outputId": "009ab71c-2d28-4993-871a-2727425a5bd4"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"sequential\"\n",
            "_________________________________________________________________\n",
            " Layer (type)                Output Shape              Param #   \n",
            "=================================================================\n",
            " dense (Dense)               (None, 128)               100480    \n",
            "                                                                 \n",
            " dense_1 (Dense)             (None, 64)                8256      \n",
            "                                                                 \n",
            " dense_2 (Dense)             (None, 10)                650       \n",
            "                                                                 \n",
            "=================================================================\n",
            "Total params: 109386 (427.29 KB)\n",
            "Trainable params: 109386 (427.29 KB)\n",
            "Non-trainable params: 0 (0.00 Byte)\n",
            "_________________________________________________________________\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "####Capire il processo il Training model.fit() in Keras"
      ],
      "metadata": {
        "id": "sPjwGZdd5NsU"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "#####Versione base sull'utilizzo di  model.fit()\n",
        "\n",
        "Questa versione base, da un idea di come lavora model.fit()\n",
        "\n",
        "Qui si lavora inserendo un elemento per volta, in pratica variando ogni volta il \"dataset\" di input in model.fit"
      ],
      "metadata": {
        "id": "2FOuNPqHNDGw"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "#Test sul modello iniziale\n",
        "print(\"Accuracy Ante-training\")\n",
        "s = model.evaluate(xtest, label_test, len(xtest))\n",
        "print(\"\\n --- \\n\")\n",
        "\n",
        "print(\"Inizio Training\")\n",
        "num_iterations = 1_000\n",
        "\n",
        "for i in range(num_iterations+1):\n",
        "\n",
        "    #Pesca random x  e y dal training_dataset\n",
        "    ri = np.random.randint(low=0, high=len(xtrain))\n",
        "    x = xtrain[ri]\n",
        "    x = tf.reshape(x, [1, -1])\n",
        "    y = label_train[ri]\n",
        "    y = tf.reshape(y, [1, -1])\n",
        "\n",
        "    if (i > 0) and ( (i % (num_iterations/5)) == 0) :\n",
        "       print(\"\\nEpoca: \",i)\n",
        "       model.fit(x=(x,),\n",
        "                 y=(y,),\n",
        "                 batch_size=1,\n",
        "                 epochs=1,\n",
        "                 verbose=2)\n",
        "\n",
        "    else:\n",
        "       model.fit(x=(x),\n",
        "                 y=(y),\n",
        "                 batch_size=1,\n",
        "                 epochs=1,\n",
        "                 verbose=0)\n",
        "\n",
        "print(\"\\nFine del training\\n\")\n",
        "\n",
        "\n",
        "#accuracy sul modello addestrato\n",
        "print(\"Accuracy Post\")\n",
        "s = model.evaluate(xtest, label_test, len(xtest))\n",
        "print(s)\n",
        "print(\"\\n --- \\n\")\n",
        "\n",
        "\n",
        "# Testiamo alcuni campioni\n",
        "print(\"Test su alcuni campioni\")\n",
        "pred = model.predict(xtest)\n",
        "pred = list(map(lambda x: np.argmax(x), pred))\n",
        "#print(pred)\n",
        "\n",
        "plt.figure(figsize=(3, 4))\n",
        "\n",
        "for i in range(16):\n",
        "    ri = np.random.randint(low=0, high=len(xtest))\n",
        "    ax = plt.subplot(4, 4, i + 1)\n",
        "    image = np.squeeze(xtest[ri])\n",
        "    image = image.reshape(28, 28)\n",
        "    plt.imshow(image)\n",
        "    plt.title(pred[ri])\n",
        "    plt.axis('off')"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 923
        },
        "id": "P3kaYaado56p",
        "outputId": "fbea32cf-edd9-4cd7-d74d-67356faeac5c"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Accuracy Ante-training\n",
            "1/1 [==============================] - 0s 188ms/step - loss: 2.4895 - categorical_accuracy: 0.0830\n",
            "\n",
            " --- \n",
            "\n",
            "Inizio Training\n",
            "\n",
            "Epoca:  200\n",
            "1/1 - 0s - loss: 0.7195 - categorical_accuracy: 1.0000 - 380ms/epoch - 380ms/step\n",
            "\n",
            "Epoca:  400\n",
            "1/1 - 0s - loss: 0.3178 - categorical_accuracy: 1.0000 - 5ms/epoch - 5ms/step\n",
            "\n",
            "Epoca:  600\n",
            "1/1 - 0s - loss: 0.2784 - categorical_accuracy: 1.0000 - 5ms/epoch - 5ms/step\n",
            "\n",
            "Epoca:  800\n",
            "1/1 - 0s - loss: 1.1461 - categorical_accuracy: 0.0000e+00 - 5ms/epoch - 5ms/step\n",
            "\n",
            "Epoca:  1000\n",
            "1/1 - 0s - loss: 0.0063 - categorical_accuracy: 1.0000 - 9ms/epoch - 9ms/step\n",
            "\n",
            "Fine del training\n",
            "\n",
            "Accuracy Post\n",
            "1/1 [==============================] - 0s 133ms/step - loss: 0.6024 - categorical_accuracy: 0.8078\n",
            "[0.6023693084716797, 0.8077999949455261]\n",
            "\n",
            " --- \n",
            "\n",
            "Test su alcuni campioni\n",
            "313/313 [==============================] - 1s 2ms/step\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 300x400 with 16 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "##### Versione più sofisticata sull'utilizzo di model.fit()  - usiamo un dataset di soli 10 elementi\n",
        "\n",
        "Qui possiamo capire meglio il lavoro in modelp.fit() di shuffle=True/False, steps_per_epoch, batch_size"
      ],
      "metadata": {
        "id": "vnZLq5sxNVal"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "# Creo un piccolo dataset di prova di soli 10 elementi\n",
        "\n",
        "xn = xtrain[:10]\n",
        "xl = label_train[:10]\n",
        "\n",
        "plt.figure(figsize=(5, 10))\n",
        "for i in range(10):\n",
        "    ax = plt.subplot(1, 10, i + 1)\n",
        "    image = np.squeeze(xn[i])\n",
        "    image = image.reshape(28, 28)\n",
        "    plt.imshow(image)\n",
        "    plt.axis('off')\n",
        "\n",
        "print(xl)"
      ],
      "metadata": {
        "id": "EwOJsuUv8vZi",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 248
        },
        "outputId": "997d67b2-cfb0-4629-83d8-306433c71e9f"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[[0. 0. 0. 0. 0. 1. 0. 0. 0. 0.]\n",
            " [1. 0. 0. 0. 0. 0. 0. 0. 0. 0.]\n",
            " [0. 0. 0. 0. 1. 0. 0. 0. 0. 0.]\n",
            " [0. 1. 0. 0. 0. 0. 0. 0. 0. 0.]\n",
            " [0. 0. 0. 0. 0. 0. 0. 0. 0. 1.]\n",
            " [0. 0. 1. 0. 0. 0. 0. 0. 0. 0.]\n",
            " [0. 1. 0. 0. 0. 0. 0. 0. 0. 0.]\n",
            " [0. 0. 0. 1. 0. 0. 0. 0. 0. 0.]\n",
            " [0. 1. 0. 0. 0. 0. 0. 0. 0. 0.]\n",
            " [0. 0. 0. 0. 1. 0. 0. 0. 0. 0.]]\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 500x1000 with 10 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "#Qui possiamo capire meglio il lavoro in modelp.fit() di shuffle=True/False, steps_per_epoch, batch_size\n",
        "# sfruttando il dataset di 10 elementi creato\n",
        "\n",
        "\n",
        "tf.config.run_functions_eagerly(True)\n",
        "#Tensorflow 2 has a config option to run functions \"eagerly\" which will enable getting Tensor values via .numpy() method.\n",
        "# serve per l'istruzione : print(tf.keras.backend.get_value(x))\n",
        "\n",
        "\n",
        "model = tf.keras.models.Sequential()\n",
        "#Layer1\n",
        "model.add(tf.keras.layers.Dense(units=64,       # units = il numero dei nodi - ovvero l'output space -\n",
        "                                  input_dim=n_layers[0],   # input_dim = dimensioni del Input Layer, oppure, Number of Input Feature model\n",
        "                                  kernel_initializer=initializers.RandomNormal(mean=0.0, stddev=0.01, seed=None),\n",
        "                                  bias_initializer=initializers.Zeros(),\n",
        "                                  activation='relu') )\n",
        "\n",
        "##Layer2\n",
        "model.add(tf.keras.layers.Dense(units=32,\n",
        "                                  input_dim=64,\n",
        "                                  kernel_initializer=initializers.RandomNormal(mean=0.0, stddev=0.01, seed=None),\n",
        "                                  bias_initializer=initializers.Zeros(),\n",
        "                                  activation='relu') )\n",
        "\n",
        "##Layer3\n",
        "model.add(tf.keras.layers.Dense(units=n_layers[3],\n",
        "                                  input_dim=32,\n",
        "                                  kernel_initializer=initializers.RandomNormal(mean=0.0, stddev=0.01, seed=None),\n",
        "                                  bias_initializer=initializers.Zeros(),\n",
        "                                  activation='softmax') )\n",
        "\n",
        "\n",
        "class printeverybatch(tf.keras.Model):\n",
        "    def train_step(self, data):\n",
        "\n",
        "        x, y = data\n",
        "\n",
        "        tf.print('\\nNew batch:')\n",
        "        tf.print('Input Picture')\n",
        "\n",
        "        #tf.print(x,summarize=-1, name='x_value')\n",
        "\n",
        "        x1 = tf.keras.backend.get_value(x[0])\n",
        "        x2 = tf.keras.backend.get_value(x[1])\n",
        "        x3 = tf.keras.backend.get_value(x[2])\n",
        "        x4 = tf.keras.backend.get_value(x[3])\n",
        "        x5 = tf.keras.backend.get_value(x[4])\n",
        "\n",
        "        xf = x1.reshape(28, 28)\n",
        "        xf2 = x2.reshape(28, 28)\n",
        "        xf3 = x3.reshape(28, 28)\n",
        "        xf4 = x4.reshape(28, 28)\n",
        "        xf5 = x5.reshape(28, 28)\n",
        "\n",
        "        plt.figure(figsize=[1.5, 1.5])\n",
        "        c = 1\n",
        "        for i in (xf,xf2,xf3,xf4,xf5):\n",
        "          ax = plt.subplot(1,5,c)\n",
        "          c= c + 1\n",
        "          image = i\n",
        "          plt.imshow(image)\n",
        "          plt.axis('off')\n",
        "\n",
        "\n",
        "        #plt.figure(figsize=[0.6, 0.6])\n",
        "        #plt.imshow(xf, cmap=plt.get_cmap('gray'))\n",
        "        #plt.imshow(xf)\n",
        "        #plt.axis('off')\n",
        "        #plt.show()\n",
        "\n",
        "        tf.print('Input Label')\n",
        "        tf.print(y,summarize=-1)\n",
        "        return super().train_step(data)\n",
        "\n",
        "\n",
        "class MyCustomCallback(tf.keras.callbacks.Callback):\n",
        "  def on_train_batch_end(self, batch, logs=None):\n",
        "    #keys = list(logs.keys())\n",
        "    print(\"------------------ batch: {} ------------------\\n\".format(batch))\n",
        "\n",
        "\n",
        "modelp=printeverybatch(inputs = model.input, outputs = model(model.input))\n",
        "\n",
        "\n",
        "\n",
        "modelp.compile(optimizer=tf.keras.optimizers.Adam(learning_rate=learning_rate),\n",
        "                loss= tf.keras.losses.CategoricalCrossentropy(from_logits=False),\n",
        "                metrics=[tf.keras.metrics.CategoricalAccuracy()])\n",
        "\n",
        "\n",
        "num_iterations = 3\n",
        "learning_rate=0.01\n",
        "my_batch_size= 5\n",
        "nr_steps_epoch = None\n",
        "\n",
        "modelp.fit(x=xn,\n",
        "                    y=xl,\n",
        "                    batch_size=my_batch_size,\n",
        "                    epochs=num_iterations,\n",
        "                    shuffle=True,\n",
        "                    steps_per_epoch = nr_steps_epoch,\n",
        "                    callbacks=[MyCustomCallback()],\n",
        "                    verbose=1) # Verbosity mode : \"auto\", 0, 1, or 2.\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "e82b1fdb-43da-4422-ddc0-ea0989b35b1f",
        "id": "PsL3kFPK2fZ-"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/3\n",
            "\n",
            "New batch:\n",
            "Input Picture\n",
            "Input Label\n",
            "[[0 1 0 0 0 0 0 0 0 0]\n",
            " [1 0 0 0 0 0 0 0 0 0]\n",
            " [0 1 0 0 0 0 0 0 0 0]\n",
            " [0 0 0 0 1 0 0 0 0 0]\n",
            " [0 0 0 0 0 0 0 0 0 1]]\n",
            "------------------ batch: 0 ------------------\n",
            "\n",
            "1/2 [==============>...............] - ETA: 0s - loss: 2.3026 - categorical_accuracy: 0.0000e+00\n",
            "New batch:\n",
            "Input Picture\n",
            "Input Label\n",
            "[[0 0 1 0 0 0 0 0 0 0]\n",
            " [0 0 0 0 0 1 0 0 0 0]\n",
            " [0 1 0 0 0 0 0 0 0 0]\n",
            " [0 0 0 1 0 0 0 0 0 0]\n",
            " [0 0 0 0 1 0 0 0 0 0]]\n",
            "------------------ batch: 1 ------------------\n",
            "\n",
            "2/2 [==============================] - 0s 159ms/step - loss: 2.3025 - categorical_accuracy: 0.1000\n",
            "Epoch 2/3\n",
            "\n",
            "New batch:\n",
            "Input Picture\n",
            "Input Label\n",
            "[[0 0 1 0 0 0 0 0 0 0]\n",
            " [0 1 0 0 0 0 0 0 0 0]\n",
            " [0 0 0 1 0 0 0 0 0 0]\n",
            " [1 0 0 0 0 0 0 0 0 0]\n",
            " [0 0 0 0 1 0 0 0 0 0]]\n",
            "------------------ batch: 0 ------------------\n",
            "\n",
            "1/2 [==============>...............] - ETA: 0s - loss: 2.2810 - categorical_accuracy: 0.4000\n",
            "New batch:\n",
            "Input Picture\n",
            "Input Label\n",
            "[[0 0 0 0 0 1 0 0 0 0]\n",
            " [0 0 0 0 1 0 0 0 0 0]\n",
            " [0 1 0 0 0 0 0 0 0 0]\n",
            " [0 0 0 0 0 0 0 0 0 1]\n",
            " [0 1 0 0 0 0 0 0 0 0]]\n",
            "------------------ batch: 1 ------------------\n",
            "\n",
            "2/2 [==============================] - 0s 151ms/step - loss: 2.2611 - categorical_accuracy: 0.4000\n",
            "Epoch 3/3\n",
            "\n",
            "New batch:\n",
            "Input Picture\n",
            "Input Label\n",
            "[[0 0 0 0 1 0 0 0 0 0]\n",
            " [0 1 0 0 0 0 0 0 0 0]\n",
            " [0 1 0 0 0 0 0 0 0 0]\n",
            " [0 0 0 0 0 0 0 0 0 1]\n",
            " [0 0 0 0 0 1 0 0 0 0]]\n",
            "------------------ batch: 0 ------------------\n",
            "\n",
            "1/2 [==============>...............] - ETA: 0s - loss: 2.1365 - categorical_accuracy: 0.6000\n",
            "New batch:\n",
            "Input Picture\n",
            "Input Label\n",
            "[[0 1 0 0 0 0 0 0 0 0]\n",
            " [0 0 0 1 0 0 0 0 0 0]\n",
            " [1 0 0 0 0 0 0 0 0 0]\n",
            " [0 0 0 0 1 0 0 0 0 0]\n",
            " [0 0 1 0 0 0 0 0 0 0]]\n",
            "------------------ batch: 1 ------------------\n",
            "\n",
            "2/2 [==============================] - 0s 164ms/step - loss: 2.1209 - categorical_accuracy: 0.5000\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<keras.src.callbacks.History at 0x7e0eafda2920>"
            ]
          },
          "metadata": {},
          "execution_count": 24
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 150x150 with 5 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 150x150 with 5 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 150x150 with 5 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 150x150 with 5 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 150x150 with 5 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 150x150 with 5 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Considerazioni dopo diverse sperimentazioni\n",
        "\n",
        "\n",
        "\n",
        "Riferimento : Dataset compoosto dai i soli 10 elementi iniziali dal dataset originale.\n",
        "\n",
        "Settando\n",
        "\n",
        "             shuffle=True\n",
        "             num_iterations = 3\n",
        "             learning_rate=0.01\n",
        "             my_batch_size= 1\n",
        "             nr_steps_epoch = None\n",
        "\n",
        "\n",
        "Keras pesca uno alla volta tutti gli elementi del dataset senza ripetizioni all'interno di ogni epoca.   L'ordine di uscita dei campioni è random, questo vale per tutte e 3 le epoche (in prarica l'ordine di uscita varia ad ogni epoca).\n",
        "\n",
        "\n",
        "$\\;$\n",
        "\n",
        "Se\n",
        "\n",
        "nr_steps_epoch > del numero dei casi nel dataset\n",
        "\n",
        "             nr_steps_epoch = 12\n",
        "\n",
        "\n",
        "Keras estrae i primi 10 elementi, tutti diversi tra loro., l'undicesimo e dodiscesimo sono ripetizoini.\n",
        "Nell'iterazione successivesi ci sono ripetizioni sparse e si conclude:\n",
        "\n",
        "6/12 [==============>...............] - ETA: 0s - loss: 2.1153 - categorical_accuracy: 0.1667    \n",
        "\n",
        "WARNING:tensorflow:Your input ran out of data; interrupting training. Make sure that your dataset or generator can generate at least `steps_per_epoch * epochs` batches (in this case, 36 batches). You may need to use the repeat() function when building your dataset\n",
        "\n",
        "$\\;$\n",
        "\n",
        "Settando\n",
        "\n",
        "\n",
        "             shuffle=True\n",
        "             num_iterations = 3\n",
        "             learning_rate=0.01\n",
        "             my_batch_size= 5\n",
        "             nr_steps_epoch = None\n",
        "\n",
        "\n",
        "Keras estrae 5 campioni per volta dal dataset, in ordine random, e sfruttando tutti i 10 elementi del dataset. Non vi sono duplicati di campioni all'interno di una singola epoca.\n",
        "\n",
        "\n",
        "\n",
        "\n",
        "In più, sfrutta la Vettorizzazione, ed il training è estremamente veloce\n",
        "\n",
        "Discorsi simile con\n",
        "\n",
        "             shuffle=True\n",
        "             num_iterations = 3\n",
        "             learning_rate=0.01\n",
        "             my_batch_size= 2\n",
        "             nr_steps_epoch = None\n",
        "\n",
        "\n",
        "\n",
        "Invece con\n",
        "\n",
        "\n",
        "\n",
        "             shuffle=True\n",
        "             num_iterations = 3\n",
        "             learning_rate=0.01\n",
        "             my_batch_size= 4\n",
        "             nr_steps_epoch = 2 o con None\n",
        "\n",
        "Keras lavora senza problemi sulla prima epoca pescando 8 elementi in oridne random, tutti diversi. All'inizio della seconda epoca:\n",
        "\n",
        "InvalidArgumentError: {{function_node __wrapped__StridedSlice_device_/job:localhost/replica:0/task:0/device:CPU:0}} slice index 2 of dimension 0 out of bounds. [Op:StridedSlice] name: strided_slice/\n",
        "\n",
        "\n",
        "$\\;$\n",
        "\n",
        "Infine\n",
        "\n",
        "             shuffle=True\n",
        "             num_iterations = 3\n",
        "             learning_rate=0.01\n",
        "             my_batch_size= 2\n",
        "             nr_steps_epoch = 4\n",
        "\n",
        "Il risultato viene mostrato nella figura successiva; ovvero nel riquadro verde i batch(2 campioni) estratti per ogni epoca (4), per un totale di 8 campioni. In rosso esecuzione di un ciclo completo sul dataset: 10 campioni.\n",
        "Nel riquadro rosso non ci sono ripetizioni; nel verde sì, ad eccezione della prima epoca.  \n",
        "\n",
        "Sembra che settando \"Shuffle = True\", dopo un primo shuffle ad inizio training, il dataset venga completamente esaurito (estratto nella sua interezza) prima di essere nuovamente ri-mescolato.\n"
      ],
      "metadata": {
        "id": "TSXiAuFK555e"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "<div>\n",
        "<img src=\"data:image/png;base64,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\", width=\"80\">\n",
        "</div>"
      ],
      "metadata": {
        "id": "7J_CYW9z71Tg"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "##### Versione più sofisticata sull'utilizzo di model.fit()  - usiamo l'intero MNIST database"
      ],
      "metadata": {
        "id": "LKl4EtOoTEql"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "# Lavorando con l'intero dataset\n",
        "# Shuffle = True\n",
        "\n",
        "tf.config.run_functions_eagerly(True)\n",
        "#Tensorflow 2 has a config option to run functions \"eagerly\" which will enable getting Tensor values via .numpy() method.\n",
        "# serve per l'istruzione : print(tf.keras.backend.get_value(x))\n",
        "\n",
        "\n",
        "model = tf.keras.models.Sequential()\n",
        "#Layer1\n",
        "model.add(tf.keras.layers.Dense(units=64,       # units = il numero dei nodi - ovvero l'output space -\n",
        "                                  input_dim=n_layers[0],   # input_dim = dimensioni del Input Layer, oppure, Number of Input Feature model\n",
        "                                  kernel_initializer=initializers.RandomNormal(mean=0.0, stddev=0.01, seed=None),\n",
        "                                  bias_initializer=initializers.Zeros(),\n",
        "                                  activation='relu') )\n",
        "\n",
        "##Layer2\n",
        "model.add(tf.keras.layers.Dense(units=32,\n",
        "                                  input_dim=64,\n",
        "                                  kernel_initializer=initializers.RandomNormal(mean=0.0, stddev=0.01, seed=None),\n",
        "                                  bias_initializer=initializers.Zeros(),\n",
        "                                  activation='relu') )\n",
        "\n",
        "##Layer3\n",
        "model.add(tf.keras.layers.Dense(units=n_layers[3],\n",
        "                                  input_dim=32,\n",
        "                                  kernel_initializer=initializers.RandomNormal(mean=0.0, stddev=0.01, seed=None),\n",
        "                                  bias_initializer=initializers.Zeros(),\n",
        "                                  activation='softmax') )\n",
        "\n",
        "\n",
        "class printeverybatch(tf.keras.Model):\n",
        "    def train_step(self, data):\n",
        "\n",
        "        x, y = data\n",
        "\n",
        "        tf.print('\\nNew batch:')\n",
        "        tf.print('Input Picture')\n",
        "\n",
        "        #tf.print(x,summarize=-1, name='x_value')\n",
        "\n",
        "        x1 = tf.keras.backend.get_value(x[0])\n",
        "        x2 = tf.keras.backend.get_value(x[1])\n",
        "        x3 = tf.keras.backend.get_value(x[2])\n",
        "\n",
        "\n",
        "        xf =  x1.reshape(28, 28)\n",
        "        xf2 = x2.reshape(28, 28)\n",
        "        xf3 = x3.reshape(28, 28)\n",
        "\n",
        "\n",
        "        plt.figure(figsize=[1.3, 1.3])\n",
        "        c = 1\n",
        "        for i in (xf,xf2,xf3):\n",
        "          ax = plt.subplot(1,3,c)\n",
        "          c= c + 1\n",
        "          image = i\n",
        "          plt.imshow(image)\n",
        "          plt.axis('off')\n",
        "\n",
        "\n",
        "        #plt.figure(figsize=[0.6, 0.6])\n",
        "        #plt.imshow(xf, cmap=plt.get_cmap('gray'))\n",
        "        #plt.imshow(xf)\n",
        "        #plt.axis('off')\n",
        "        #plt.show()\n",
        "\n",
        "        tf.print('\\nInput Label')\n",
        "        tf.print(y,summarize=-1)\n",
        "        return super().train_step(data)\n",
        "\n",
        "\n",
        "class MyCustomCallback(tf.keras.callbacks.Callback):\n",
        "  def on_train_batch_end(self, batch, logs=None):\n",
        "    #keys = list(logs.keys())\n",
        "    print(\"------------------ batch: {} ------------------\\n\".format(batch))\n",
        "\n",
        "\n",
        "modelp=printeverybatch(inputs = model.input, outputs = model(model.input))\n",
        "\n",
        "\n",
        "\n",
        "modelp.compile(optimizer=tf.keras.optimizers.Adam(learning_rate=learning_rate),\n",
        "                loss= tf.keras.losses.CategoricalCrossentropy(from_logits=False),\n",
        "                metrics=[tf.keras.metrics.CategoricalAccuracy()])\n",
        "\n",
        "\n",
        "num_iterations = 2\n",
        "learning_rate=0.01\n",
        "my_batch_size= 3\n",
        "nr_steps_epoch = 4\n",
        "\n",
        "modelp.fit(x=xtrain,\n",
        "                    y=label_train,\n",
        "                    batch_size=my_batch_size,\n",
        "                    epochs=num_iterations,\n",
        "                    shuffle=True,\n",
        "                    steps_per_epoch = nr_steps_epoch,\n",
        "                    callbacks=[MyCustomCallback()],\n",
        "                    verbose=2) # Verbosity mode : \"auto\", 0, 1, or 2.\n",
        "\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "8ae5ffe9-c77a-4417-c5d9-831a85844386",
        "id": "9g9idLkulugD"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/2\n",
            "\n",
            "New batch:\n",
            "Input Picture\n",
            "\n",
            "Input Label\n",
            "[[0 0 0 0 1 0 0 0 0 0]\n",
            " [0 0 1 0 0 0 0 0 0 0]\n",
            " [1 0 0 0 0 0 0 0 0 0]]\n",
            "------------------ batch: 0 ------------------\n",
            "\n",
            "\n",
            "New batch:\n",
            "Input Picture\n",
            "\n",
            "Input Label\n",
            "[[0 0 0 1 0 0 0 0 0 0]\n",
            " [0 0 1 0 0 0 0 0 0 0]\n",
            " [0 0 0 0 0 0 0 0 1 0]]\n",
            "------------------ batch: 1 ------------------\n",
            "\n",
            "\n",
            "New batch:\n",
            "Input Picture\n",
            "\n",
            "Input Label\n",
            "[[0 1 0 0 0 0 0 0 0 0]\n",
            " [0 0 0 0 0 0 0 1 0 0]\n",
            " [1 0 0 0 0 0 0 0 0 0]]\n",
            "------------------ batch: 2 ------------------\n",
            "\n",
            "\n",
            "New batch:\n",
            "Input Picture\n",
            "\n",
            "Input Label\n",
            "[[0 0 0 0 0 0 0 0 1 0]\n",
            " [0 0 0 0 0 1 0 0 0 0]\n",
            " [0 0 1 0 0 0 0 0 0 0]]\n",
            "------------------ batch: 3 ------------------\n",
            "\n",
            "4/4 - 1s - loss: 2.2999 - categorical_accuracy: 0.2500 - 606ms/epoch - 152ms/step\n",
            "Epoch 2/2\n",
            "\n",
            "New batch:\n",
            "Input Picture\n",
            "\n",
            "Input Label\n",
            "[[0 1 0 0 0 0 0 0 0 0]\n",
            " [0 0 0 0 1 0 0 0 0 0]\n",
            " [1 0 0 0 0 0 0 0 0 0]]\n",
            "------------------ batch: 0 ------------------\n",
            "\n",
            "\n",
            "New batch:\n",
            "Input Picture\n",
            "\n",
            "Input Label\n",
            "[[0 0 0 1 0 0 0 0 0 0]\n",
            " [0 0 1 0 0 0 0 0 0 0]\n",
            " [0 0 0 0 0 0 0 0 0 1]]\n",
            "------------------ batch: 1 ------------------\n",
            "\n",
            "\n",
            "New batch:\n",
            "Input Picture\n",
            "\n",
            "Input Label\n",
            "[[0 1 0 0 0 0 0 0 0 0]\n",
            " [0 0 0 0 0 0 0 0 1 0]\n",
            " [0 1 0 0 0 0 0 0 0 0]]\n",
            "------------------ batch: 2 ------------------\n",
            "\n",
            "\n",
            "New batch:\n",
            "Input Picture\n",
            "\n",
            "Input Label\n",
            "[[0 1 0 0 0 0 0 0 0 0]\n",
            " [0 0 0 0 0 0 1 0 0 0]\n",
            " [1 0 0 0 0 0 0 0 0 0]]\n",
            "------------------ batch: 3 ------------------\n",
            "\n",
            "4/4 - 0s - loss: 2.3099 - categorical_accuracy: 0.0833 - 433ms/epoch - 108ms/step\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<keras.src.callbacks.History at 0x7e0eaecb4970>"
            ]
          },
          "metadata": {},
          "execution_count": 48
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 130x130 with 3 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 130x130 with 3 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 130x130 with 3 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 130x130 with 3 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 130x130 with 3 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 130x130 with 3 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 130x130 with 3 Axes>"
            ],
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAHgAAAAxCAYAAAARM212AAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjcuMSwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/bCgiHAAAACXBIWXMAAA9hAAAPYQGoP6dpAAAIDklEQVR4nO2ceWwU1x2Avzm86117fWLjtcEXGBvTQIKBmIYGCLQgjqBAnRCKgEIRtCRFSSPa0BZaVWnTqEipqrgh5UhIELSFQAIKbt2CSjAYMEeCjTGHHWMb3wa8eL3rnaN/0DqcBVUzC1nN99fu0+r9fnrfvDe/efO0gq7rOhYhi/igE7AwF0twiGMJDnEswSGOJTjEsQSHOJbgEMcSHOLI9/vDb4oFZuZxV4q1v96x/WHK52HK5VasGRziWIJDHEtwiHPf92Aj8E0fxaXne8hyt9CxPpW43ZWoV64GM4XbEOx2xNQUGqYm0ZmjkD6gmdF9aihuyMG2KQ7Xh2XoimJafM/sfFSbQEcuZK05j9raamj/QRMsulykrTzD75OKSZYVvpHzCvElMfCABAthNrqmPUbDBMj92kW+n/gRWfYmXKKPJMnPgrhDLFk8By5kw7EK0/J4YsVhYmUvAV1ib8kYnHu9aF1dhvUfNMGeSbkUJq+hn+zgohK43igIwQp/E+LQHC5OjWPI1Cp+mnSAZPkqb7eO442ySeAJIye3jvUD/sLqzF0sLlhC5jHzclmZUEKUGE6j6qV4WQ5CdT+oqDKs/+AIHvUI9qWNxEkSfj3A7FMLSTymoV8O7uwVZJnuycO5OEtl+ahPGOesYu6JhfhqXMSWC6TX9dAyQiJlxFVaNZkt7Y/jbDLvIpRions/uyUndllBF2yGxgiK4LpJLspy1mMX7FQEemB7PK59p4N3/xUEpLhY2qdkY5/bxLqBH3OoK4uZ//oBA9cqyGfPQmw0zROSSJ9cQ0H8EVZUz6JpZxopW86hmpGTKNE+PRe7UAzAqZ4Abbv7kdJ8ztAwpguWBmaQOr6WMEEC4LTfTewZb1CLK8Fmwz8sgydfKmVp/KfMr5yHpyiJwdsvojQ0IqT1o26GmydnH2Na7ElerXwG++Y43B+fRPV6zclJkrg6/Rp24bqCP7aOo++RLjSDx8VcwaJE4yQ3JYPeRAxuwX5zGq5I6iba2JV4mGfPFxDxi0giSg+iAHJGGhcWpDDz6QOMc1Wy/MRsktaHY/vbYTQTTzOJkRFERfh6vx9syCC1thUl0GNoHFNHXRAFfPH0zl6A2p4+CKpmZtjb84hykT++gg6th/a1aUSVliKE2RDTUqj8YSJ/mLKBCKGHRQfnk75RRN5XBmYeVRMllNw0lmQW9TZFO3zozvDrhaeBsYO60aGg8qeiiciNl4MZFgIK5a1uJEAJFxAjItBGDkZc5+OfM9ZQ5Uvm5d8tIfuVBuS9x8yVy/Xl2ZMazoKoS71tO4e8T8vYvkgxMYbGMn/dvOESOuSzk7X5Kkp9g+lhb0S/1oW/NJW4PDtPvHCUooyRvPWddxht72Zp3RTqVw+i7/4TqD7fvTsziVjRgWoTQDS2ajd3Bgsizry23q+Lty9BqG82fYbciubx0L/Yw+ttw1iTdISKRW8xLjzAi/VP0bw8Ddu+z9EeoFwzMU2wEGbDO+VRdgzbgIiAJIhIPoIutxdN43BH+k1NFR1JyK2d6AYXNv8PCiqiqoNm7PiYJliMiWbu67txSw4ASnwaAwqrUds7zAp5RwRZhmHZXHutmx2DPuJ9TxIjf7WMgguT2PPIJk6v6oM0MCOoOd2JUUfnkfRJHeplY+sTcwQLAsS4SJQ7AdDQWbh5GVqnx5Rwd03DbkcYkkX1t6PYmruJDzr7s37lMyRuPI5/jo0R215m3ZPvciWvL6LTGbzERAElXKBd6+5t8nQ6IBAwPJQpRZac7ObsL6OY6GgDwgAYWFiDYtKmwR3JH0rV/HBWj99Jpq2FFXXTaflZBpGHPkPz+9GjIpg05iQBXcbRGkDr7r53nwYhOp0kzKvlnct5vBp/Gr+uEH0kHN1rfA6GCxZdLq7m9+fNUe9hF8J627uHpNDz9TRcReWGvi25FSk+jtYZ2Tiea2JL1kYAvndyHv1eA7m8AsHpRBg8gLpVOqviS5izaxmDqy+hBLM2kESGx9axKKYMcBJAJa7Sj9ZtfKFn+BIt2G10uSXGhl+5qX3obz+jI0dCsBm7mX4jclp/zv8om6dePMS6nA9Ilrt5o34yrm0u2h6NouW7w7m2NYb+b3/B5sc2MGfXMga960FtbDYtpzuiKPzjUjZuORKAgK4RXt2GrnwVlmhBRLVx0+wFWJ24n7KqPHQTl8L2MSnkj6/g5wmlOAQHnZqPyQkVrJnhZtrAcuyiwo8TDhMp2Jl+diZZm7ug/Fzwq2hZZqz7PKr+5Y6e7u025QkjaBvEw/csJ7ekFsXvNy2G1KMTJmg4/vPKLUoMZ67rCybkF5Iu/7eIsvOT5jyurE0luvxztAfxiCRJPB19PCihTN+qPNGjkXd0Lrm/bkZpMneTI7q8gwO1mVxUvizm7ILcK9evK7zU+Dh/f280UTtOoAWz6LsLLaqXgjNzwKQL3/gZrKnIXp3Lmo8mVWLpqfn0f6EzONuTbVdwfprALMcins8sY6SjBp8exoWeRBr8sRxoGUD31iSSt51CM3EluSeqys4reeT3PcY2zxDE3/RBu2bO+BguWO/yElflZ965Z6m7HEPCBidKvXFHUP4XamsriYUdSEWp/HnctyjMUxF9IpG1Iq56FdeechxdNQT3Xdbt6N5uiraORntO4MMjI8jef9y0g33C/f6Fw8N2ev9hyudhyuVWrHPRIY4lOMSxBIc4930PtvhqYs3gEMcSHOJYgkMcS3CIYwkOcSzBIY4lOMSxBIc4luAQ59+USOh5TUmjEQAAAABJRU5ErkJggg==\n"
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 130x130 with 3 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "#Nota\n",
        "#Il dataset dopo del training non cambia, shuffle=True riguarda solo model.fit()\n",
        "\n",
        "print(\"\\nDataset dopo del training\\n\")\n",
        "plt.figure(figsize=(10, 20))\n",
        "for i in range(30):\n",
        "    ax = plt.subplot(1, 30, i + 1)\n",
        "    image = np.squeeze(xtrain[i])\n",
        "    image = image.reshape(28, 28)\n",
        "    plt.imshow(image)\n",
        "    plt.axis('off')"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 112
        },
        "id": "hrUdcFWbLahw",
        "outputId": "0cd6fd79-28e8-46b3-ec9f-856936140e43"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\n",
            "Dataset dopo del training\n",
            "\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 1000x2000 with 30 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Considerazioni dopo diverse sperimentazioni\n",
        "\n",
        "Riferimento : l'intero MNIST database\n",
        "\n",
        "$\\;$\n",
        "\n",
        "\"Shuffle = False\"\n",
        "\n",
        "Settando i parametri\n",
        "\n",
        "\n",
        "         num_iterations = 3\n",
        "         learning_rate=0.01\n",
        "         my_batch_size= 1\n",
        "         nr_steps_epoch = 5\n",
        "         Shuffle = False\n",
        "\n",
        "\n",
        "\n",
        "L'estrazione dei 15 campioni restituisce sempre i primi 15 elementi del dataset in input, elaborati -con processo ottimizzazione- uno per volta (my_batch_size= 1); 5 per ogni epoca.\n",
        "\n",
        "\n",
        "Eseguendo il codice più volte, si osserva che i 15 campioni estratti rimangono sempre gli stessi, e seguono l'ordine presente nel dataset\n",
        "\n",
        "\n",
        "Settando i parametri\n",
        "\n",
        "\n",
        "         num_iterations = 2\n",
        "         learning_rate=0.01\n",
        "         my_batch_size= 3\n",
        "         nr_steps_epoch = 5\n",
        "         Shuffle = False\n",
        "\n",
        "\n",
        "\n",
        "L'estrazione dei 30 campioni restituisce sempre i primi 30 elementi del dataset in input, con lo stesso identico ordine.\n",
        "\n",
        "Eseguendo il codice più volte, si osserva che i 30 campioni estratti rimangono sempre gli stessi, e seguono l'ordine presente nel dataset. Vedi figura sotto:"
      ],
      "metadata": {
        "id": "ZI5EtEGKe_Ms"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "<div>\n",
        "<img src=\"data:image/png;base64,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\n",
        "\", width=\"200\">\n",
        "</div>"
      ],
      "metadata": {
        "id": "ZN0i_6C2-aN1"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "# Nota l'orinde di estrazione con Shuffle = False, nel training precedente,\n",
        "# è esattamente lo stesso ordine del dataset dato in input\n",
        "plt.figure(figsize=(10, 20))\n",
        "for i in range(30):\n",
        "    ax = plt.subplot(1, 30, i + 1)\n",
        "    image = np.squeeze(xtrain[i])\n",
        "    image = image.reshape(28, 28)\n",
        "    plt.imshow(image)\n",
        "    plt.axis('off')"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 58
        },
        "id": "9unc0qpAIFaf",
        "outputId": "1304f975-68fc-4820-bfd5-965ead81e272"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 1000x2000 with 30 Axes>"
            ],
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAxsAAAApCAYAAABdhDTyAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjcuMSwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/bCgiHAAAACXBIWXMAAA9hAAAPYQGoP6dpAAA5LElEQVR4nO2dd3QV1fqwn5k5Pb33XkhC710UARGVqhQb2AUr9u7FepUqiopgFxRFEUFBRAFBSiihJ0AoIb33cnLOzHx/HEAsKORMvPf+vnnWYi1IDvudObNn77dvQVVVFR0dHR0dHR0dHR0dHY0R/9MXoKOjo6Ojo6Ojo6PzfxPd2NDR0dHR0dHR0dHRaRV0Y0NHR0dHR0dHR0dHp1XQjQ0dHR0dHR0dHR0dnVZBNzZ0dHR0dHR0dHR0dFoF3djQ0dHR0dHR0dHR0WkVdGNDR0dHR0dHR0dHR6dV0I0NHR0dHR0dHR0dHZ1WwXC+HxwsXtMqF/CD8oUuR5ejy9Hl6HJ0ObocXc7/N3L+L92LLkeX83fokQ0dHR2d/1JEi4W6sb0omdLnP30pOjo6Ojo6LUI3NnR0dHT+CzFEhHP45U7c9txXxI0/8p++HJ3/IKKXF+W39Oboos4YIsL/05ej838UQ1Qk2bN6UX1dLwSj6T99OTr/h9De2BAEEAQEownRwwPJ2xvBaEIKDEAKCUa02TQTJQX4UzS1D212GMn+uDP0aK/Z2H+JILjuy2xuNRGizYYUEozk7e36Tv9XOT0XNHzup6m4qTcD9jaSu7QdhtAQzcc/gyghWiyu53H2n6AgzeeAYDC4nvl/CMFoIntOLw6/39WtcQwxUZRO7o3wUwRHPuhKznO9Uft0/J+Yy4LZjGixIHp5udYtPz8kb+8zf0SbrXXvQxCQ2iTiscTOmtEzaFYNVE+L1lyMaLEgBQb8TzyTv0WUXPtNgP+v7+Y/rCyd3ue0XhOkkGAOT2vLFff+jNwkoTY1aTq+5pznfDLEx3Lkwy4YoiJb71pECaVfJ/rsaUZaF47YMRVEqfXk/QOINptrnrXC/JZD/XjxiiVMeXopYkyEpmOLFguOId0oubsPRxd1pmuGwvGXetM8tHur6Aetjii59oSzdALRywvBbHY9mxauq6ef7290jQB/13ckSv+z8/e8azbOhWixIFjMEB6C089GTZwFp1mgNhZsHSoJ9KynaHU7nrz1U9LMhVz79gNE/Huz2xcuBQZw5JFktkyYjkWQeCZkPdeFj0Mc7oFSX39hYwUFASCXlp7X5w3RkWS9GIjvzxYCF6aDIl/w9f/l+KEhZL4awaFLF9D+l0nE31OMXFyiqYzWRvTwQO6URE2chfIOAnKonZRX6pEPHNJmfC8vyno7ud1vFyFtq/kiajAUFWsy9mmkwAAau8ZRlWSiqp2TrKvmIZ5ln2c6HFz92VQSZ2af99z5KwSzmYK7uiINqCBkTAOq0+n2mBeC5O1Nzl3tmHvle9y37Ca3xjn4WBi7r5qFj2iFFNfP27a9jpjH45EPH9XoirVHSk0i684AVG8H8VGlPBv/Dd9Wd+JAdRgAiiqQXRiM93orgbvrkA6fRK6q1vQaDHExVMxRWRq7irvyLufwzDS8Nu1G1VKIIHDikS68MXE+r153HWzdq+Xo/yiStzdlo9tiH1nFoKhDjPPfhkOVuHnxXcS/tAeloaF1BAsCoqcngsEAskzJuLY89NBnzHppPH4fbQXV/Scm+fqQ+XwMsXGFbLu1C8k7diJrMG5rIaUmUdYjkMDNxchHjv3lZ2s6hbCg30ImT7mdhGllKBobUYLZTN1VnQi+9xj3+u+g3i+dKXOvpvGpDogbMzQUJGCIjKC+fRhOm4h3ZpVm+9wfECWyXmvLwoHv8cDrdxD+9i5Nv7eqNh6M8SzjpLORz8yXuj2eaLEghoVQnxZMzlUC8we9zwBrw5l9dNrEDNaNtXBX+nVEfGLEIyMXZ2GR23J/j2A24+jXDtPWrAvWD88gSkgJMTgDvUCAqiQbgZNyWJS4FG/RAsCQzJGc2BeOoV4gaq0d6ec9F6QfSolxZD3lz3v936e/5df9f1FtMNN2XIXvzxYEBTzznUh2BfOubM33n9bCLWND8vPjxJRU7CmNdIzJo6dfBjf57sbv1BcP0GvXBOoSnPxrz5U4czxJXlGOW6q5ILjSC+6JZvW46fiIVrbYJR49NIankr7loQdvJm5+9gUp58WjE1EkgeA3z09hlAO98fFuQGq2/P2HW0D5pXEs7jcPBQVV1c7rKIUEU9M/jvI0iWY/Bf99Ar5HmrD7G/HclI1cXuG2DNFmw94nlRNXGbn5kvUM8doHQKyhmZ6O+0id6o1cU+O2HKWuDu9MI3mD3baXz8nR+5O5Y+T3XO65nyiDiIgRBeXM79sYJd665h0eO3Q7/h9WuG10ilYLk25eTbHDm/2hETjz8t29hQvC2T6e6yb8SJNiIu6blm9gcm0tkWsEll4cxyTvAgrlBiINnmzr+S5db5hK7NP/jLHh8jT709g+kvz+RgwNAtErKlD2Z53z/2RPDGTr6BkEiFYa1WY2NfnQ2ZZDZ1sOuQ5/ggy1pMXmE9XPzryK3mx4rg+2r7Zpds2ixULBsHDeaPMGhx0qW75vT/y6Q8h2u2YyTuPwVog3VFPc3ZOQrZoP/wckXx8a+ibT5CdR3sG1rkWtaca4dqdb42Y/1pZRl29hWVZH1n7SizVSL1JHHuLr62cy8ciD+H24RYvLB1xOlKZ+qVQlmXBaoS7BieTlwCPdimlwGf5SHeYaRRNDA1Ei75a2XN55F5lPt0fdvt39Mc8lysODuiHtKOopophVwjeqLZrXRRcH8vYjcxn3wxTSpjXhzC8452ct5c1YBAf9L9lH0Ws+KEXaKM2C0YSYGMORmwLpd9F+ngpbxRGnGQ/ByWcJ39Cv3X0EbRK0eUaCQP3oHhSPsRPiX060VyUZBZHE3aLNPvd71F7tmDFgCd3MddT3aEBc4ouikXIuGE2U9FIxChI3Zt6IT4n7+oDaLpEjD0o81XkZE7yKKZMbOeQwwCktMEh0cokVDg54l2Xd/Jn2/nVEzahAdTS3TKAoYQgORKmq/o0RVjOqM70fTidzfCz8jRF8zqE7tKHyZTtDw7cjCQqdbScYYq0HzCinXEGrU5dBKtQpdm4bMJyjS3oSMn/Hed9P4dAwMgbNxCaYztI0YIJXMRMuWUhmPweKKlAie1IhezJt8QTi3sjSRHdrbdzT1EQB//5FrGn3GUbBFdpRMDGnMpkoYwUDrLn4z/Ag9FgBOByoTe57AdVeHSh/up6v02YRaXCFq+MNdVTsCqY0zpsxozeyY1UnuABjwzSyhPKM4PP7sCBQ0c6bSyO3siu3i/ZRjZgoKoc3EGNoZMLRUUS8a0LRwHI1xEZz8NFQpl3yJSnmQjY3JOF7eQNVso3CZl82vdQTzy/cU5oMsdEcfDyEe/r+yPU++xCBgTtvxfyNL68/9QaD2mWSHxICGizCgsFIsxfYxNbz/jf7y9zld4gCp8JVmRMor3eFejuF5DMrchVeool+libqIwQCRAFV+ZsBz5Noczn7rPHaDPYnSEnxHJ0UQsKsQ2cWKcFo4uSlNjracpi69Cbit+5suSddVfFce5AXhl7FiGFz2GEPJdJQh6dowecfCmrIF3fh2BgjA7ofpKf3KkZ7HSbL4cEDJZMJ2H/u/xexzsFFzQ9hLREQm1W8TzqRmlzvuGiXUQ0iJd2sfH3/q0wL2kPikN4kf6XddattE2gzIYtVNR1ZO6Mv8SsPtI7nSlUJ26yyb0QwddEqrZiEiJScQEWPIIovcfJUnxWEGqsYYKlCRuXZS/uTObk9pO9r0dhC17a8cPViHvt+PKkvnsBZVAyixMnC7ux7NoyUOw9Q/KGGNxMXhWNqBfPbLMGEQnpTHCbBiW+PBtoYS3jy5Ag8j1SjxVLQeFVXRt24gXXP9MW6xj2D7K8wxMeS+WAIzw9ayjBbLjbRyIdDY5gfNILA+RduqHkJDpYOnse1pfeR8ErdnyrdgsHAyUEWogwNpHgWkpMyAEmDyLQhKpITN0TTY/g+Poj4AJsgIWJi0OopeAQ1kN7jfbdlnEGUqLq+B13vyWDt2s54znRwtG0Ivabsp9ikfYqTITKCnIecXGYr4ZOaRCI+NeEsdj+ifpqiyd347IrXACMFef54N7k/drO/hQDfSoZ55DCrohMfLR5M0B4HpzeXwj4GvrtxOtEGK6M8Kng0qRlBElEdLZNXM7475cMbiH8lEDIOAGAIC8VxfQXLD3UgKffcjqa/Q7YZuSpiJw8HHMShyrxe2ZavykKYEbEGT/G3qZOeoplF8auYeXsuP//S7cy1/B3tr9uPTTj33Ek1GgFoSxPQROzE15nS9VoC/x2FsHlPi+/tTxEEDOFhqB5WBHszztwCt/Rdt4wNubwCw9wE2l55N5KPg+0XzeOFkn5kjY5ANRl5bnwosQeP4NQgxQRcaS2HxlnZ0v5NmlWVbum3YDE6mRS/heAdCi+YR/PIFd+w1dPI+Wa1CQYDnQPzWcv5GRuSry897t7F8foAzAU17kVpfj92gD+Zj4TzQ6+Z5MpmSufEY/shHdVND4whJorMh8L47vLZPHxiDHXTI/E4UMzBJ0PYNnQOn1S3x1LhvtJ+9OYIdg2biVEQKZVVRu++lah7alGtzdSrJjwMdhC1KRMSjAbsgTJegkqCqYTCvl5EHPPX1MJPffIII96dhOBUsJ4sIsLumsflIUG8+00n7vc/yFtVSYRttmua8mQRHKgerRM1AzhySwhfjp/N1B/vwvCT6/uSggN59rpPeeHIFSS9U4izpZ6lUyi1tSR+4mRY5CS+7/Ah4DLUBG1t898gmM1UTOhC8KQTPBO9kFrFSpVs4197ruSddcMJ+/o4gRW7/tKIMq3ZSeyPEqqigvpHD7UxLJTasTFEGqxst6tErNUw8ti2DZd8uJV4UynTp12Lz6fbkTV2ZpyN97ZccpqDGDvoF/ZERP6lF/pCEIwm6JCEZVYZw4L3EWHMIsVYxur6VA42hBPlU06TKuMlmpjk/wtT/TvTUtXM6WVmVvZgEpY2uwwNAEUm4Oc8Pi/uztiQ7XwY1luT1AwpKIi8FwS+T/uIYRm3Yt/uT9yifBqTgrjl9WV4iAoH1ycSm+V+BEK02cgdrvBzSSK273ajttI8kAIDOPhoEJ8OfpP5JRfzwuJxNPsqLBs1hzmDaxE+MKNeQFTN+6STPfYIRnmW8MjoZbyfMQKPL9P/8B5JQYFE98kjRDJjEZzIZvG89+xzYYiJ4vDLAfzSfzpmQcQmuNZQBQVLgZF6bGecolpQN6Ybox9ey5J5g4h/dyeqLFMxtgcHykPxK9O2oYPk7U325Gi+6DKbJlXm1R+uInlNhibzQjCaqL66C/dPWUqaUQaM7l/wKSxbD2O93Ztro6ZgKKsj6lj6mb1StNkIldpRfr2ZaCDbYcecZ0SVW26q14+txrDDD3XPr47T+k5RTIpfzaJpV7iVcmY4nMe6u/qwwdETAKm8DkFW6HXTg/gfVBny2Ebu8k8/k9kjIpJmzWdN6EWcbxXXzhXtmHiFRFZ5MMJKf8zVKoUXK/TucISxwduJNZTjLzkIk6wAdDXDpi4f033A/USlG/5SD5H8/FBiw1H/xPARDAZEH28I8KNwSAjBo07iVEWGh2cw1COTFXXt+PLpy9yK4rut+Zm/207KQ/uJeU9kYXV7vtrbGTm/EPnIMaKe36xJLjucyie/rQ0bRs+gVlEZkn4n0VMqsC7wZeWEfti+TickHa71OsaxkabzLtKTe7fnUp+D530dgqcHF/lkkXEiCorKWno7f34tSZE8NnAFgZLE+I13uB6sm4aG2CGFo9N9WX3VLMZm3Ioy2RPzt9upTw3hgT5rEIE3fhyC6Zfzs7z/imYflZdK+9J/5ySGfvQw4XdU4szNozHeHy9R23xcpaGBhC+b2dgUQV+Lg47j9qNGhWoqQ66sRN15AGVPJnJlJUpDA4LVQsEVEQzwcHlIvs7vhOVEuWYyGxQTUcZyapK8NBvzbKSQYJSIJqoUC5L91EYlShSMjKWnJZfiYl/U2hbmtP4OU3Yh1rf92G73OfOzissbNRn7DwgCpZO6MODerdhlAzesmMLMSRP4eEg/Ym84TNBbW3AWFv294qSqrgVbkX/z7okeHqh9OlLwti8Zo+awxS4xeeY9eH6tjcdZMJo4Ns6f+/0OsyCvP75ZtWe8SIb4WJT+nV3FrVoiu8bv5ZlNXWftinTtAzvQfsFBFiesYKJ3DvWKmaknrubb4d3IujKYfz11C983RNOkOnnm5AisuS2PdBo27cV3VC7iht/m4Ku1tZys8aOXJZ+iq+LcvaVTgyo0NRkplo04N/kTM2MXziBvSu5spIs5lyt33Ubc0kpNHA/NvVMZ2DYLx/zQlqeUnAfVlySycNB73Js5nsJ+jURP20zK3EI21LchOajU1UDgAjB/t4NX54ynWLZzvXcujRMrkfz9/vA51eGgpNYThyqzuzYK255ct+7DEBXJ4ZcD+LnfG+Q6jfTZdisPFvbCocq8Xx2L72GXAutQtTHapLRkHJMqmP/zQILf3YnqaMY+tAv3X7WS0mKfvx/gAnG2i+Or62aRajTyr6KBtFlQdUFG4G8QBFfTi9Qkmi/rxslHunHNk2uokm0M3HODptct19TgzM1D2LwH+fDRM++GlJrE8cc6ct30lbQzqthVB6O330HC/JwWz3fRZuO6hB2YK/iNB74u3EB7Sy4Gu5vxRknCWFLrqm/buhc5+zhKcSlhW52oEnyytwf1imvPUFD4vC6Y52bdgOWH868PinxpM+V9KwkafojAd7bgtWQryZPTKe9byWv3TODOp+/jsncfoe/u8XxeF4xddWAUJCIHnfxbnVeuqoJ9f15LVHxnD469FUHi4pP88Oh0BoVkkrclghk/X45ZgKu99lPRxj1DXZOEd6WhAUt2CQv29ePlPl/xQfth5x02Oi9EicIb2jHlhhV8VNWVj74eSOhWJ86iYqzLi34TsjYLRkb1SyfT7Hleec4lXa1EGcsxV56fhzLnumi6mvMRis2aFh4awkI5dI/ETd65PFh4EUlvtDCOeBZC57ZUvmhnVduFDEufTNwzTchZ2Uh+fuQNkhjnfZAJh8eTOqcQpwZFZikvHGVPYgfCsguQyw7hPKWsVcUb8RLcv5/fI27IILspFDz+gXzFU7m5BSMc3N1lNe1MDipkJ4XpYcSeSNdEhFJXzze57bkk9SB2bxFPTUY9C1GibGgCs3t9yMQNt5Cy9zAKrnSN4bdv4KWiy0h6s9l9B4EoYb+8C/m3NDAyfguXWu2c9mv4/GR1+zZ+L0vp24GcYRYEJ6xY2YuEDwpJOuYqQtAi3iQYTZSN60DvKTtYHLIem2DhxeNXErqhAkWjiJboYSWgRzEiAtlFQSSfzIcAf2ouTiJ/iMKEHtvIa/Ilc0FvAj7eqakCKgkKqkYOX8nXh5wrJD4O3IiCgaeL+rDhzZ6ErCtEPnYcgMo28fS35rCn2ZPcz+MJzmy5t0x1OuFPnoFcVU1pfjJyO7D7aRN9Uqqq8fk+maNdgrj6+vWszrsIy82FfNdmEVftuo2IZ1SUvS1P0ziNYDRR3N3MyWOJJP+SgxOXkwBfb+QjxzVN3RUUcKgGqvYFEmgrprlLIicHWCh2HCPniwRCyi7QmFZVwpYd4/Xb+vNCSDoftf+Qqcl3Imz5NV2zYVgnqmMN3Jr0rSaRBik5gYOP+bGr/1yOOU2M3XQHKU+UsnpKN4xDZNYt7EnIN3tpCO7I4SHu12mIFgvZ1wcQIJaQMrcc2dHsSn++q4qulhPEL9K4gF+UyLncRorRzA+NVnbO7oz3/pYXWYk2G4XXpRIyOodxoZsJNVbz6K7RhCy2IPhI0FnDa/8T5Iu7ID1bxGexn9DWZKBOcTJg5y0ELbLiLHQjlU4QaFKMKEbX3087ixpCBSQUbHkNqKLU4vdHLi75TXq+0K0dWbdYmXbxVySbiokxNBJ4KuLwaW0Ec+ZdTdgn+zTbI0yrt2MCfEQJ0cPGs8+MZeC46ZglI0W1XoSrf6MHnXak/Qkjb1vPBN/t3JM9jp4/P0jkOoX4nEqypnogAU8UXE7Ummq3mpRoVl3rzMkl/lVvDn0Qhu21Eg5s6E38ksq/LMY8X6T4aNped5C25nyefuo24lbvQXU4/5BedLqWOthUS6Z4fp5hpw2aVCOB+85Shk+1LBMk10IomM0obeM5NtqT6WM+4of6NkT+KLfcs/B7BIGSy+P4od901jQGsmlhN4J3bXfrwQoGAydGebOq7XQmHJhE/MPVyPlFSGnJNMy2szjpdeZV9IB/BeI8rk1nDrm0FKG0FMVgoOqGXtSHub5HqXclFkFhxcH2JOdkaiLrNAqt17ZT6NqW7PHemOJqEUWVVzp8SB9zBTbRSIbdyM3vP0DC0lLNUl1Up5OyMi/k1jj+RhCwD+3C8AfX8VNNGslvNaPU1kKP9oQ+fJTxPtu5dv5DBKe73ylO9LBxcojEzu4L8JNsnB1AreiocGG+0r9G6JqG9YVCEhURYZKEUlaBs6XdRs6B6OsDo8uZGbYVEddm8nriEh57cyRZa/oQ+1UZctZRt5TA5k4JvJq8kEK5AemYFbWpibrBacQ/lMmCsNUA5Dp92PNAPivKL8X6zXZtClw1xpkWy8QBG/EVDdx5cig509sQuHInTkczgsGA2i2NyItyMQJTD4wlZFtN66QIiRIYXe0cDGcH0wQBBLFFz0p1OvHOaeZQUxj3B+yk5/NH6Wep5rHCSwl40wNl7w5NLl0K8MPYuwJpvT9yWQUFj/RhyLitBBhz+OyDSwmbvUWzZ9/sKZJkLKf/xfs41jWAO6KX8fjm0aRP6ULorl0oLdjjlJpavtzXmWnB20g0Giju6UFkQRSll0RSGy3w8g0fESpVk2pqxqEK7C4NJ6Sp5Upm7sgQll0ymyNOIzfPv4+UD4/hLC4l5rsgDnzVlpADro5kAZnNHHacZ23mXyAGBdKh/xEOLU/G+8g2BLOZnPFRfJI2m6s330nihj2ado7Lf6Qnz4xdwgFHMw++ezfRX+1yqyZIaWwibHUhleVRvGuJxivfQUJGDnJ5Bc3XdgdANMsIGrfEloKCKLsikQH3buWFkHREDFQrTVx/ZCzhz4uw1710QaW+no8P9GDEjVvZWNWLgK/2ozbZUY0QKjVw4hGB2OlpqNtbVh92NoLBQMFFXmy7YvpZDZFce0OZ3Mjs+VcT/t6elne++isUGUESGT94E/6SmZ+bTFi+8kVpbLmuvfztAfyU2w/P/UUkl+xF8PIk65k4fho4kxX1yeQ+noSU4V5NiKatfNSMA6x7oi+XvPQLcycu45Y+42me3h3z2owWh5YFo4nj14UxJ+ITRqTfSfRX2/7yRRMRqHZaQTnP11115chXtDESVu46p6OskyeNQQL2dg2YTE7uStvAiaYG8o6n4Ss2sLymM7ZdOdrUa4gSdVd3586Hl9Gkijzw2U3Ef7THfWtYkrAHyYRIJhzLg8gfLhAyQuHB2KV0MdViFESu/aE/STv3alLMCK7FpOaieApGOPii/2tYBJkYg4BRkFhaF0PSXKfm7Q1bC7FdCpd99Au3+xwGQBIEDEgop7Iv5xQOJvYNV3qV1gSIjdgDNKwF8PYm77Z2LL/3VcINZr5v8OG7qW2JfqcLJ640sThiNVesnErSW9p07VFqa2mzoIoBMbfwROpqjIJMgrGUTmYzgfHaRqFKO3vyU8I7dP76flLqjrTK4q7W1qJ+nchg0xiGh+8l0lROL0sDXyR8j3Pyd6ya6MeT799IzDuHkMsuPKVO8vMjt5+FIKmR8QdvJPr7Jpr6phD3cCb3ha5lbMat+H7kRd5lKm8M+ohllv/es1iNOaUs+nYAnyZ2Jf7JemxHtv2qeHVOJWBmLgtjvuf6Y6MIva361zoLLRElDLFRdE44iVEAa6mC2DGVplAPSrqYcNpUEt8vxHnsxAUPLTU6KXd44CmYuchSyxPF/dg1ozPe6/66FuhCkCMC6RRykqKfBMQAPwIH57P3/o7IVomLX9rOztwebjfzOI21XOalwqH08cnmzah15DnteO03I6a3fA9SGhpImufgiXZ9mB66jV8emoX8kIqEay/4tX24q1JnUvw2lnUb1KKuZFJyAkrPamIMMtf3GkVEwZYzEXVxYwYq/Oa5SBo8JWeEP8meu6nZEonk403JmBSm3fIJN+2dSJunK3Fq2bK8VwdGjt/IKI9CRh8eQ8zyMmR391BFRs4+jnf28TM/Oq3LWMtkDjQ3clFSNiUeHpo0cwHXfpp1uw/fDp+Jh6jwaW08O2rjWPVLZ9o8fxi5XJvOi8kvNPDLjHhefuYdXr95EGWNQUyLX0S+7IlxhxfqDm32OFWW8cpVeKWkP909j+MrNXCp1ZXp4i+ZSRp9mIY1kaBxC2TBaEJISyDrASsfBXxHhawyOf1Wktblnpn3LSHobVerbicuPe74GyH80nMmS2vbsvyuS5HW73L72jXvG2r+djuby7vxwe19eH/Aeyya1puD3t3w/LxlYT8pPARLlwpsgorPco9zfk60WKhMETnsaOLztX1JtJ/flxO028G/c4dxzS0/kXNtAKKgMNLzJF8XdeJIfjDyIS8+Wn4l/vuqiUYlf4kfh6uC8CxuWfu0P9xfUhwBd53gRu98vm0IJnSbrEl6lupwErBTYuWlQWx8+jXWNXkzN2cQt22cxL96LSfYUEvYL6o2qWCCgJQUT+bjvszut4h5Jy/h6tV3I9pFvhk5m0QjtDfnc+gWK2lFkThz89yX+TvCLDUUeZg0i3OIlTW8tnkwm1MTOFIehKIKiIJKhE81L8Uu48XIb7it+/0Y12jjzTwjt9REjEGgLl6bDUvy9ib39nZMv/NdKhQTI7bfSNfQPDb3e5Nt3QPoYipDBixFEpKPt2adj5T9WURNjWFhzChUUSB/gJlFN84h1qeCWi8vV1TFDQyhIchl5QTsa2DQ7oksuHwhk+tuJ/GlA5q3nFSamghYuAXpc2/WBnekOcqPx0eYeHjICm7yOcFVthpib5vNxPqphM1Lv2DHiqNdLLeOX02B0wvrC944vAxEPn2Eu0N/5Jpv7iVhqZ1jowVeGfgZd/90A2mb89zaWH5PvWLGWKtNdMGZX0DcE65C87NHFAwGyp+181XM98ypaE/RvAS8ijTotytKSAH+OBPDqUmw0uwp4PAWqEtp5rWwNYRJViY+vYIjjSGUNntSUxZK5XE/ZJ9z7yXnFOXlRfYIGy/4bwMkcpxO1n7eg4glmzX1ZFeleuHtNCHUN6EG+NL4oR8+G7ciAuvb98ExthbPL7SRZVmZTtHOUD7oPYKr587h2fwrifixym1nl3Q0n5WH2vFK6BaMgnSm5DjPaSdf9uTb6k6M8d1ORxPEm4upDzPhe4EyRJuNzEf82NB9DmbB5Mo0OMd7IRgMlKeZSDC6X0NameKBKKjkX2SjYWIiSwe/RlZzGF7ve+M85n4mx2mkkGCybrCywH8L6XZvyhZFE3ik9bqSAZhLG8hqDmHjsQSSG4///X/4O0QJKSiA0n8rHOr8JpkOkQHf3U/CZ06k9btIZKumTXbkg4cJuC2cxwffTk08ODxVDoRF8uSqa0l8YzeKVuumquK9ci8HjqSyx6cTjUFGXrillAUpn5BoNPN45HfcOPx+oo9atHOwCgJ1IzoTPfUw66NWYBNMjD06goRXHO7rVKe+F0NYKFmPxrKj1yymlVzM1tnd8P1FG0dK6xxSsG0fqceDeH7RlaxIWcpVtwdg2NJCJVMQMBud5DhtBKw7+ac52ILBQPFNnXl8wufcnHkDye+Xn3dfevOq7TQWprE2IvbMzw7L7bAeLSc5/9AZZVwBCh/oQ1tTAaUZIXjivrEhGE0UDgpmRuQy7KqDZw5cRcSP+7WJNCgyQUv280bZOF4NlLCVyXgeKCMhVCFlQCEflPXHI69Bk0mkXNSJiofr2dDhNQZuvIeklxtJKz/JsblBeIkyDxZcjFl0sHzoXK4NvgW/TyLwPFyNWF2nmeFxu/8mRnfrQ9i2v+7IcL448wtIfayJutAgwkrKOd3XVg4N4oW3r+CFqG/clvFn+B0QkFERbE5Ei/sLlbNtHL2u2cMzh4YjLg4gav1JjnVO5dGnVN6MWoeIFQWF6IE5qLO1bSPsPJ6DdDwHAHP7PviKzXTyzuMX36QWGxuC0UTlhK40j6nEe2EMlm93EvhkMic+C+LOK79n7YddNfPG/R65pgZqapCyj9Nmrz9LVg/jyIs7eTV0B+1NRmxDixEWmi94/jUFmQg3VvLiiSsw7j9O9owkPolcyUUrHiR1Vj5HJkeybNRsRi27n7TXCjQ31gscfliOlmhS33IuKq/tzoZOr2EUJN77fiBJyzPcW+cEAXlAZ46NMZLcNo9ufhlc7JWJSZCREbjI0oyIwMScS8lYkYbXSQXPgmZCswoIrs67YCeLFOBP/o0pfDFuNs2I3FPQh/uDf0TW9rBwAJo9BQrqfPAqLkVtbMQ3+1d/fMSPVRiHV6DlqSuuTl0xGJHYXRRBdI42XclOI6sqe5slHsseQ+3SMKwVCoZGhbKnPXkzal2LxxWsFm7psYlA0cT4o8NR7eeOagppiYy8aQNeogODm/61wG1lrDjRjl4j9rLuUDJFsjdPrR9N6qaj2inOokTF4Hj+PfhT/EUDA7dOJPEjbWu1zoWMgFJuRnW4vyIInVM4+ojEivZvcdghMurrqaRNP6lZ57s/w5lfgN8HBfjh6lC2oV0SlnIBpUnbs4qUhgbIOIAEeALCifbMeH0Ib0ato4NJwtGxDsHLCzQwNpR+nci+3sSk3ht5InAf1YpAp4/uI3yjE/Nubc7fMYSGkPlUDN9eMYs5Fd3JeLYLvt/v0mzOaWtsnDpRtW5wGnnDZGbHfIaCQm6ZL/FF7ln8r+QOQ6n5nYIiCEiJcWTfEsJTI7/gX2vGkPrSCZwlF9YlStl9EPPu3/7szxYNhzfUq0ZiVmtjqYqxkbww9T16mut5vrQXkQ/bkTUsOldqa7F+nc7pklwZcCR1J80os+5EEnGHNEgF69UBj+cKeDF6JUM+eJjkT8tQPC3kzAtgfqePGfTLXcS/6kRwyIwf1Yu+V+5h5Ks7qZI9ePabsSQ87J7i9N6GATwwai+RBjOelxchLvJpUSrLbxAlUBVXG92zW+kKAsSHE2yuc2/8v8BSreBQFawedgRPD7cXKkN1I7vndyD4h1ycuVtxAtYwf8Is1SytC2X2jLHYSmU8j1Qj17sXyhYtFhDFPyhyUmIccaOPkmD05N2fLiExv+WLo31gB6Y++RnP7x+Gx7FqCAqgtKsvEcZK/p0xlKR8bSKOZxAERKsVtbn5N0aEXF6BaXUFX43uzqtXuKJbpZVe+MoX/h2erjUrrPYmxlPB5NWMSRAIjK2gxzdHSZPzmPLgfbTZdAznBZwfdL44VAlV4434bESbjfKhrsKJ23MvJv6rBreNaKVvR3rN2c407z08dXQUn//Uhx8O9MUz30GTv4E3XppLkexN8aNxRG399VCtFqlPokTlkGReu/ttPq/qzk8z+lAfJnLdlM2Ytc+iBMCpiL/OubPmXXknH8T6ZjTtdyQINHuIZDsV7Ie9QXavMxSAUlOH1yYbw0NGcfhEKAmfKNj252Iuc6VrGKIiKbe7okvtTWVUtQE/o6lFio0kCOw7GkmK448H6AhmM0JqAh6vl3Cj7zYGrZlK2prjbkUG5axsIm4LpMjDH/FuE7/UJtPm7Qb3952zcAzsxA1PfMsIjzK6pd9M8gP5yFobGqKE5OeDvWMcTg8Jya4gOFWm7bsS1aiipsVTm+BBZRuR+IUnLthAkIKCyJrkyfreM7EIAj1X3XvG0BA9XM9eiAyjId6Pmsk1xPpWoKgi+3LDafNiHXKm++2DHRH+9AvexQ/FYWh2ENY5kIoqyaoKRo5U2dBkwesnDxQ306xFi4Wim7sQM+4oP8V+SYhk4pv6QB5dcS3JMw8hV2izAEm+PmQ9GscPV85ARmDT1F5Y1u/UtJ5OM2NDCgqitl88VTfW8lnnOSQbTRTLjbxU2gfrNk+3raMXYr7mwU6TMe47BmYzzoQwqpJsxN52mIXhXzFx9e0k37utVb1zp1E1Kpyq6RjEEGs9ChIrP+9DZLZ2J92eC499hcioOB2S2xEA0cuLymca+Cx+Gd03TsHvBGQ96s2svkuoV8w8+MJkEpdluVrIAlH7IW9ZCg+O7YBigDZzj7n9vMI2ChQPbybSYKVP8HEO+MVBSxd9QUBKiKVocChBGfWuFnenf2U0IfdMw/piEf8O20COU0SQtS/Q9fh2N9Of7kvX8FxKoqJafi+nkA8exv/gr0qWFBTEkWu8eNp7D9d/fyfJC1xzzt0lRQoJ5tDj8SheTtKeK0YpLgVJQgwK4Pi4MNITZrHTLhL/pd2tQmrb4VKe3D6S2T2X0Gd1KaWyQIHsxcyTl5H4UpOmKVSCwUDT4M7kXKOStMCBsOWsAjlRQkqMpX97V05updKIR7oNpfnCu6557ytja10C73X+gOvuuZdrkn/BUzCysdNiMuwityy8h+hvd7mfq30OvMQmCPC9oINQzxdDZAQnbozhvV5v8khhf7LmtsVn+w63I6r1ERau9tnBrS/dT8C76SQoOSAIqH06Yn2sgKzmMGbOHkvwVg28wT3actEjWzlgj2DXre3xc1SRek8+tYoVj2LtFRhjA5gkGdHb+zed4aS0ZNSry3F+GQRkayNMEGge0hXfSblMybqWxE+rNHmHVEczwW9uhjchGZdD6TdvvaLQ6HQlV4VIZtTEekR/X1fHn/PF4WRrRRyy/16yLnuLwUPvwlJix1hQiSPCH6eHgbxLjLw/bh5HmkMZvPxBUp/LxumuUaCqyKWlNPTryeNXfM1bM0YRkKHt3l3WwcztPicolO1YVvggF59/e/6/5VTac/ElwfiPzWN5m3mIiFQozeyw/9o+/ki/ELzEJhYc7Yv6me3CZIgSJcMTefPy9wiTrBTLjRg8HZQOjsFSFUn5DfUYJIVx8bu433/fb7qSdf3mHuRMDZrWCAIVbW0UNvngt6/GrfPKDGGh1HeKAsB2pAxKylFlGWQZwWSiuUsih68XWZs6B6NgpUqxYalU3NaxBC8vDMPK+CLxOxTMTC3oT8bsTiR+sUsz41Mwmsif1JZ3h7/N8toOLJ59GQHrtmregMRtY0P08EDukEjmLUYe77uC671OYFdVHivqzjdrexK/rJ7Qre53uEk0iMTPzOKH7R0QfZu5uf1mhnntZWl1N27/ZDKpSyo0zf07F0ZBpjHY5HZbUkNYKEkPuRaQT2sjCN127nxTLSm7OApJo8oGwWLhueTlmAUDi3svwNFLYlVNR57YMxKv5V74L0r/Q6cmZX8WMaccUFoYhl7LM3j3id48G7Sbf4fs5NLEbpiPtMy7LXp6cvxlTz7tNofRK+4lLLrXmWdSlSjR9opDLIhdya5mK5MzriP2aKn2xq0ss/JYWxrKbSTv1rYeBFGiamACT4/4gnmFl5L6tIZhf38fApLKmZv6Gbf63YgjMxLZrBKYVsbHqXNoUGXGfn0/bfZluiXTeTyHNs8bean7jVQliaiiiqlGIOrbMpRMbYvxnP060O65vTzgu4/50y7FeaqdohQSTP74RPyuKOC1yNUomHkg73LC11WitMSQKq3g2x+702vEUQ5e/wYKCu9UJfFVQScK0sNJ/DRfk9bUv0dtaGRvXSRX+WdQl+SLVUN9Bk51bBkZw0e3zcGhShx6IA3vjVs1Sd2UmlWaVAM1CaDe1gPPfJm6CInet+4i3lrKK/PHEbl4vyYbcn2kla4ex5mVPZiA/DIOPRzH5IBv+Pexy/HZV675vhOwo5zKCSI5tyYR854Avt5UdA+kZmQdzr1eJHy6R7umHskJ8FApg0IyWXPXRSh7tOlM+HcolVUUrOrEllgzvS123uq+iOe73oz5u/M3NuS6emrnpvLI4/15MWw9D0xfzMHGCFbkt+OaqJ+JMZUhIzAz7zKK30gg5acjmkUfRC8vCq+x83FuL4I+067JymnU/lUA3Ht8DEFbtZ1jhvAwDj7sz8bLpuMvmnikcADfbeuEqVzCVqhiLf/t3QTkN6GcuMDMFEUmeEMRk/tdz08DXyPSYGXPRe9Q2LeZZlUk0Xh2/qHL0Pi8LphFBT3xzdZmVxUkiYp+djblxBN/+Khb607ZoDhmTZuHhMqTx0ZxPD8Rao0Yq0WaQx3c2WMDH/tkEChZsasOFuRehLlSg/sI8uPddh9x+qDFOqeJ4p7Q7NkVQQZzjYLvphzk0rIWGzZiUiyv3/smexpjWPLKZQR+vsvtg6T/jBYbG4LRRM2YLlSOqeeGNltZ7LcLH9HCu9VxzPliBDGr6kjcu8ftAmS1qprSQ6kUt21mbsTPVIetxYhAkQxXbrqLhHkK8ZkHNSts/TssgkxNjPtnIDjiQ7kvZD7FsoOXvxhD/DbtF6w/w6PYgYiIILg/mZSqap54+VaqL23EaHJiW+WF/8EG4o4X4SzJ1LQf/LlQHU4+/akvT4/bhbtnVApmE091+I5Uk8jHV75F/tBfD6NKMRUTY1B5t6otH747lNhleThPnHTz6v8cUVQJiqjCEBGOM0+bLh0AhogwOj20m0KHH/kvJ2Ep0+Z8EAAlOwf5m65cW3Yb6QNfp6GbihEIM3jiUCU6b7uJlJm5ODV4T+VD2fgcyv5NKklrzDSHp4EeXkfpbynjg4+byDzeGbVZwj+0mgXtXqOzSUTEwivlqZx4JQXbwd0tkiNXVpL06mEWfjeKef6uTcVabMdaXE188X6cbhbTn1NuTQ17ypK4M3gdtVEGND4BBTqnYh5WgoTKjQvuIXKTdt5fr03HuOmDe5h09Y90sx3jWHMw22viWLWpMwlLm4jI2K1ZSqrTLOAh2rEaHRQt8GFJu7lcvfYuUl6rRc7UriD4NMrhY4jvdSXuruOEjazH21BGZk0D9qWxxC7O1Oy+RIuF4+OD+SRxDuO+vJeEjdrkfp8PSkMDkfP3ca96J8vufZV+FihvZyR89QWcg6DIeKzIIKuiHZ2u7cD2y+cwxFrB/f6u1qbrmry55/uJtFlQi+fe7Zq1KAeoH5jK+LQt/PRyX8z1JzQbF8AQF8Pz7b7h87pgCt6Lx1/jonDVbgdZ4IOqbry37mLavF9Dm8P7/lJXa4m2IGcfJ/5jf26NupblKZ9jFoxEG35dZYrlRgZsuAf/nyyITpduYsmpwnZUm4JkISWR53stZ9qysW53KZTN0MPsuqrVqcsgFeoUOxWKQqTBfKrDmpUyuZF9zX44p4di2Zbltk5X2jOAKMk1iojIu9HrKI74jmNXeaKoIuWyJ+uqU/kpJ4XwN0y0pGuUYjXSpBiZ/eNQ2nyR0aJ21+dDi4yNxhE9CHzwOP+KnEeisYlSWWB/sxd375lA5EsCsXtcObJaKM9yVTUprxxnRO4jTLl1Ob5SA9MPD8brLR+S1x9AaWj4RyIaAOIpJ5nzAiOKf4qq0oyIQwVbkdA6/Zj/BPOmg8yrSiU8oBohIhRaGAUAV6g84P10ghYZXbn6jY1n2qf9YygyUT/KPHVxD7p6nMDQ5Eaf7qpq5r44lm/vyGJO9Eo6m5zkOJ2sqmvH4sJeHPwlnsRFFYQdSte2veHZSBIPpq7l5T1DCWzULgcYQG1s5PtDqSg1RlK3ahjVwDUXgt7bScgSKwMmP4zdX2VA/30M89/Lgrz+RD/l0NRw+ifw2H6C57++huhr5vNi9HKIBg9RoUkVcKgi6xq9ufP7m0lc1IR1y3a3vEFyWTnS+nLOXlr+iXWtam8gTakGHBfenOkvEYwm8p+QWdXuAy7bfgexH+do2kFLLi0l5qVKNr0ezibp1OnnTidJjRmodrumjhuPIgc/16awJGUxuU4jk3ZPInV2NfLBwxpK+RXV6cTzyx0oP3iSb7CQjwWc9QTXbNNOYRYE6oZ1ZOHEN7h63RTSZp7A+Q84h85Gqa0l4u09XOH9CM6kBhLX1V5wjrjqaEZav4uUrRYuufthLp2QjoTCqmW9CNtiJ3nTbs2VJykwgLIbGli8oS9tvjuguZOwtkMIsYZyPq3qiW92o+ZF4XJZOSkPNLLZFEJyY+splwCG9buRDgRw0aip1PRv4oVuX/PUzhF4b7RiqlFJ+TnPtS+cWhu0nIH1id54iHb897m/7vhlNdJz57V82fFdIk8ZTJ6imXq1kR12id1NMczYehmRKyW8fs7GVL5Dk85XQSuz6ZHyIBdfvJfxgVsJEBuIMYj0NssoOFjbaCLAVEe70EJqnrGgrr9wGWpGFrNHjyGlILvVUnWhhcZGQ7DE8coAJldeR1WFB747zfgcdxC5Zk+rdEtwFhUTNquYZbOCAAjEtcj/E5GAs4lcW8uotMkEH3ZfsiHzJNd+eQ/GmHpCTmh/uva5UBob+fj9y3ji9k95cso42jxX6d5ZEYqM4oaCrwXmb7ez91vYSwwSLe8HrTqd+H68lcq9qfS6+QFMoQ2Yt3gRtr4adU8WcUpxqyuAqsPJv5dcjUehilqvXbMAcG0yide7DJjWuA/V0Yxc1Uzky660yTzgHeKB/H/MIaAlcnEJCU9V8OymW6mOM+DwAnuAgrVYxFyuEri3juSMnZp0P/tPkfRGDjfLdxG31r3TYX+P/dKODIvdzkvFgwj8wAO52P1iz9+jOp2tcs7N7zFu2s/Wp3vQe3xnzJlWYr8q1aR49S9R5FaN1ktpyfjdl8P84otJfM95qiPVP49SX0/Ms671wp35pzQ1ET5jM5kzXP+Owv0xz4m/L6MT9/Djl33dbuH9Z1iXpzNq2N1YCozEZWW1ytqpNDSAttvLOQTJyMUlBL1dQtDb8D4xJLD7zK9bc+Us7iox7+Ql+H7ufqRE3LSbsBMRDLn7YTxTf11zGvb5Eb7JiamiieTtO131PG7KOhu5tJT4R0spiIzg+Y430+QrUdxPwTe8BrvDQMBiD7yyKhFq6mlOCcVICxx6ioyyV/sI7e9pkbERsGALLHD9/ezzOP/7zrPVmPR9JF6vzVByZSUJD2nQZ/5CUVWiPj/JE9HjuGPQj6z9uh/ihtbfsP9nUFWU3QdJuvesH/2T8hX5zOb7TxvTOn9EdTqxrEzH8me/+8evRnuc+QXEPlWg+b00+UkM8d7HA6/fQfgPu1D+gZadrYVqt2NZmU7iSte//xcN57ORQoLJvM+bkHqZyjkx2H7RLp3y/wtKK1gzux9B64+3mrKcfIcrpe1/fa79J4le00RTRjg2h/vd1QCcefnEP/afic478/Ix5+VjBnwW/fZ3p+eI8b88c6B1ztnQ+a/GmZtHynMNrPukF8Yjx/QFTUdHR1N8Vx5geva1hGft0+5QKx1NUKqqSZlbi1hrR87f2SrFoP+XkSsr8ftwyz+bLqxzwYgbM9Ai411HGwRVX2l0dHR0dHR0dHR0dFoB99r36Ojo6Ojo6Ojo6OjonAPd2NDR0dHR0dHR0dHRaRV0Y0NHR0dHR0dHR0dHp1XQjQ0dHR0dHR0dHR0dnVZBNzZ0dHR0dHR0dHR0dFoF3djQ0dHR0dHR0dHR0WkVdGNDR0dHR0dHR0dHR6dV0I0NHR0dHR0dHR0dHZ1WQTc2dHR0dHR0dHR0dHRahf8HDStoVXLXjvgAAAAASUVORK5CYII=\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "\n",
        "$\\;$\n",
        "\n",
        "\"Shuffle = True\":\n",
        "\n",
        "\n",
        "Settando i parametri\n",
        "\n",
        "\n",
        "         num_iterations = 3\n",
        "         learning_rate=0.01\n",
        "         my_batch_size= 1\n",
        "         nr_steps_epoch = 5\n",
        "         Shuffle = True\n",
        "\n",
        "\n",
        "Ogni epoca è composta da 5 campioni, per ognuno (my_batch_size= 1) dei campioni si attiva il processo ottimizzazione. Ci sono un totale di 3 iterazioni o epoche.\n",
        "\n",
        "Tutti i 15 campioni estratti sono differenti tra loro.\n",
        "\n",
        "\n",
        "\n",
        "\n",
        "\n",
        "Settando i parametri\n",
        "\n",
        "\n",
        "         num_iterations = 3\n",
        "         learning_rate=0.01\n",
        "         my_batch_size= 3\n",
        "         nr_steps_epoch = 5\n",
        "         Shuffle = True\n",
        "\n",
        "\n",
        "\n",
        "Ogni epoca è composta da 15 campioni (che sono tutti diversi tra loro, non ci sono ripetizioni), per ogni 3 (my_batch_size= 3) campioni si attiva il processo ottimizzazione. Ci sono un totale di 3 iterazioni o epoche.\n",
        "\n",
        "Tutti i 45 campioni estratti sono differenti tra loro.\n",
        "\n"
      ],
      "metadata": {
        "id": "qpRfxG8dcR9a"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "####L'uso di Callback objects in custom training loop\n",
        "\n",
        "\n",
        "https://github.com/keras-team/keras/blob/v3.1.1/keras/callbacks/callback.py#L88-L89\n"
      ],
      "metadata": {
        "id": "N3hzaSkvoUpJ"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "class MyCallback(tf.keras.callbacks.Callback):\n",
        "\n",
        "    def __init__(self, data, patience=0):\n",
        "        self.data = data\n",
        "        super(MyCallback, self).__init__()\n",
        "\n",
        "    def on_train_batch_begin(self, batch, logs=None):\n",
        "        if np.random.random() > 0.993 :\n",
        "           print(\"\\n\")\n",
        "           print(\"Info su output dell'ultima layer : \",model.layers[2].output)\n",
        "           print(\"self.params = \",self.params)\n",
        "           print(\"data.shape : \",self.data.shape )\n",
        "           print(\"\\n\")\n"
      ],
      "metadata": {
        "id": "1SQ35rZAoTzd"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "####Fase di Training"
      ],
      "metadata": {
        "id": "KQkHLU4-Jd1l"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "num_iterations = 20\n",
        "my_batch_size = 8\n",
        "nr_step_epoch = 120\n",
        "\n",
        "\n",
        "#Training\n",
        "history = model.fit(x=xtrain,\n",
        "                    y=label_train,\n",
        "                    batch_size=my_batch_size,\n",
        "                    epochs=num_iterations,\n",
        "                    shuffle=True,\n",
        "                    steps_per_epoch = nr_step_epoch,\n",
        "                    callbacks=[MyCallback(xtrain)],\n",
        "                    validation_split=0.1,\n",
        "                    verbose=\"auto\") # Verbosity mode : \"auto\", 0, 1, or 2.\n",
        "\n",
        " #\n",
        " # NOTE sui parametri inseriti in model.fit()\n",
        " #\n",
        " # epochs =\tInteger.\n",
        " # Number of epochs to train the model. An epoch is an iteration over the entire x and y data provided\n",
        " # (unless the steps_per_epoch flag is set to something other than None).\n",
        " # In pratica, a meno che non vi siano diverse specifiche in steps_per_epoch, per ogni epoca si intende l'intero dataset immesso in x e y).\n",
        "\n",
        " # steps_per_epoch = Integer or None.\n",
        " # Total number of steps (batches of samples) before declaring one epoch finished and starting the next epoch.\n",
        " # When training with input tensors such as backend-native tensors, the default None is equal to the\n",
        " # number of samples in your dataset divided by the batch size.\n",
        " # If steps_per_epoch is None, the epoch will run until the input dataset is exhausted.\n",
        " # If steps_per_epoch=-1 the training will run indefinitely with an infinitely repeating dataset.\n",
        "\n",
        " # shuffle =  Boolean,\n",
        " # Shuffle argument in model.fit is set to True (which is the default), the training data will be randomly shuffled at each epoch.\n",
        " # Has no effect when steps_per_epoch is not None.\n",
        " #https://faroit.com/keras-docs/2.0.1/getting-started/faq/#is-the-data-shuffled-during-training\n",
        " #Alternativa:\n",
        " #https://www.tensorflow.org/guide/data?hl=it\n",
        "\n",
        " #validation_split = Float between 0 and 1.\n",
        " # Fraction of the training data to be used as validation data.\n",
        " #The model will set apart this fraction of the training data, will not train on it,\n",
        " #and will evaluate the loss and any model metrics on this data at the end of each epoch. The validation data is selected from the last samples in the x and y data provided, before shuffling.\n",
        " #https://keras.io/api/models/model_training_apis/\n",
        "\n",
        " #https://faroit.com/keras-docs/2.1.2/models/sequential/"
      ],
      "metadata": {
        "id": "rpE2u7Ej5K3H",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "52ae1a7c-7b7f-4cc4-b793-1f6305f07a16"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/20\n",
            "120/120 [==============================] - 4s 26ms/step - loss: 1.3337 - categorical_accuracy: 0.5979 - val_loss: 0.6060 - val_categorical_accuracy: 0.8368\n",
            "Epoch 2/20\n",
            " 31/120 [======>.......................] - ETA: 0s - loss: 0.6757 - categorical_accuracy: 0.7903\n",
            "\n",
            "Info su output dell'ultima layer :  KerasTensor(type_spec=TensorSpec(shape=(None, 10), dtype=tf.float32, name=None), name='dense_2/Softmax:0', description=\"created by layer 'dense_2'\")\n",
            "self.params =  {'verbose': 1, 'epochs': 20, 'steps': 120}\n",
            "data.shape :  (60000, 784)\n",
            "\n",
            "\n",
            "120/120 [==============================] - 2s 14ms/step - loss: 0.5491 - categorical_accuracy: 0.8354 - val_loss: 0.4032 - val_categorical_accuracy: 0.8843\n",
            "Epoch 3/20\n",
            "120/120 [==============================] - 1s 11ms/step - loss: 0.4701 - categorical_accuracy: 0.8542 - val_loss: 0.3347 - val_categorical_accuracy: 0.9043\n",
            "Epoch 4/20\n",
            "120/120 [==============================] - 1s 11ms/step - loss: 0.3406 - categorical_accuracy: 0.9042 - val_loss: 0.2678 - val_categorical_accuracy: 0.9203\n",
            "Epoch 5/20\n",
            "120/120 [==============================] - 2s 13ms/step - loss: 0.4335 - categorical_accuracy: 0.8875 - val_loss: 0.3026 - val_categorical_accuracy: 0.9123\n",
            "Epoch 6/20\n",
            "  1/120 [..............................] - ETA: 0s - loss: 0.0889 - categorical_accuracy: 1.0000\n",
            "\n",
            "Info su output dell'ultima layer :  KerasTensor(type_spec=TensorSpec(shape=(None, 10), dtype=tf.float32, name=None), name='dense_2/Softmax:0', description=\"created by layer 'dense_2'\")\n",
            "self.params =  {'verbose': 1, 'epochs': 20, 'steps': 120}\n",
            "data.shape :  (60000, 784)\n",
            "\n",
            "\n",
            "120/120 [==============================] - 2s 14ms/step - loss: 0.3329 - categorical_accuracy: 0.8938 - val_loss: 0.2841 - val_categorical_accuracy: 0.9160\n",
            "Epoch 7/20\n",
            "  1/120 [..............................] - ETA: 1s - loss: 0.8649 - categorical_accuracy: 0.8750\n",
            "\n",
            "Info su output dell'ultima layer :  KerasTensor(type_spec=TensorSpec(shape=(None, 10), dtype=tf.float32, name=None), name='dense_2/Softmax:0', description=\"created by layer 'dense_2'\")\n",
            "self.params =  {'verbose': 1, 'epochs': 20, 'steps': 120}\n",
            "data.shape :  (60000, 784)\n",
            "\n",
            "\n",
            "120/120 [==============================] - 2s 13ms/step - loss: 0.3636 - categorical_accuracy: 0.8938 - val_loss: 0.2467 - val_categorical_accuracy: 0.9273\n",
            "Epoch 8/20\n",
            "  1/120 [..............................] - ETA: 0s - loss: 0.0435 - categorical_accuracy: 1.0000\n",
            "\n",
            "Info su output dell'ultima layer :  KerasTensor(type_spec=TensorSpec(shape=(None, 10), dtype=tf.float32, name=None), name='dense_2/Softmax:0', description=\"created by layer 'dense_2'\")\n",
            "self.params =  {'verbose': 1, 'epochs': 20, 'steps': 120}\n",
            "data.shape :  (60000, 784)\n",
            "\n",
            "\n",
            " 39/120 [========>.....................] - ETA: 0s - loss: 0.3066 - categorical_accuracy: 0.9038\n",
            "\n",
            "Info su output dell'ultima layer :  KerasTensor(type_spec=TensorSpec(shape=(None, 10), dtype=tf.float32, name=None), name='dense_2/Softmax:0', description=\"created by layer 'dense_2'\")\n",
            "self.params =  {'verbose': 1, 'epochs': 20, 'steps': 120}\n",
            "data.shape :  (60000, 784)\n",
            "\n",
            "\n",
            "120/120 [==============================] - 3s 24ms/step - loss: 0.3066 - categorical_accuracy: 0.9010 - val_loss: 0.2341 - val_categorical_accuracy: 0.9312\n",
            "Epoch 9/20\n",
            "120/120 [==============================] - 2s 14ms/step - loss: 0.3197 - categorical_accuracy: 0.9010 - val_loss: 0.2317 - val_categorical_accuracy: 0.9302\n",
            "Epoch 10/20\n",
            "  1/120 [..............................] - ETA: 1s - loss: 0.3080 - categorical_accuracy: 0.8750\n",
            "\n",
            "Info su output dell'ultima layer :  KerasTensor(type_spec=TensorSpec(shape=(None, 10), dtype=tf.float32, name=None), name='dense_2/Softmax:0', description=\"created by layer 'dense_2'\")\n",
            "self.params =  {'verbose': 1, 'epochs': 20, 'steps': 120}\n",
            "data.shape :  (60000, 784)\n",
            "\n",
            "\n",
            "120/120 [==============================] - 2s 13ms/step - loss: 0.2939 - categorical_accuracy: 0.9094 - val_loss: 0.1930 - val_categorical_accuracy: 0.9467\n",
            "Epoch 11/20\n",
            "117/120 [============================>.] - ETA: 0s - loss: 0.2365 - categorical_accuracy: 0.9274\n",
            "\n",
            "Info su output dell'ultima layer :  KerasTensor(type_spec=TensorSpec(shape=(None, 10), dtype=tf.float32, name=None), name='dense_2/Softmax:0', description=\"created by layer 'dense_2'\")\n",
            "self.params =  {'verbose': 1, 'epochs': 20, 'steps': 120}\n",
            "data.shape :  (60000, 784)\n",
            "\n",
            "\n",
            "120/120 [==============================] - 2s 13ms/step - loss: 0.2329 - categorical_accuracy: 0.9292 - val_loss: 0.1951 - val_categorical_accuracy: 0.9450\n",
            "Epoch 12/20\n",
            "120/120 [==============================] - 2s 13ms/step - loss: 0.2465 - categorical_accuracy: 0.9229 - val_loss: 0.1964 - val_categorical_accuracy: 0.9418\n",
            "Epoch 13/20\n",
            "120/120 [==============================] - 1s 11ms/step - loss: 0.2586 - categorical_accuracy: 0.9208 - val_loss: 0.1809 - val_categorical_accuracy: 0.9477\n",
            "Epoch 14/20\n",
            " 39/120 [========>.....................] - ETA: 0s - loss: 0.2584 - categorical_accuracy: 0.9231\n",
            "\n",
            "Info su output dell'ultima layer :  KerasTensor(type_spec=TensorSpec(shape=(None, 10), dtype=tf.float32, name=None), name='dense_2/Softmax:0', description=\"created by layer 'dense_2'\")\n",
            "self.params =  {'verbose': 1, 'epochs': 20, 'steps': 120}\n",
            "data.shape :  (60000, 784)\n",
            "\n",
            "\n",
            "120/120 [==============================] - 1s 11ms/step - loss: 0.2151 - categorical_accuracy: 0.9344 - val_loss: 0.1893 - val_categorical_accuracy: 0.9447\n",
            "Epoch 15/20\n",
            " 21/120 [====>.........................] - ETA: 0s - loss: 0.3270 - categorical_accuracy: 0.8869\n",
            "\n",
            "Info su output dell'ultima layer :  KerasTensor(type_spec=TensorSpec(shape=(None, 10), dtype=tf.float32, name=None), name='dense_2/Softmax:0', description=\"created by layer 'dense_2'\")\n",
            "self.params =  {'verbose': 1, 'epochs': 20, 'steps': 120}\n",
            "data.shape :  (60000, 784)\n",
            "\n",
            "\n",
            " 41/120 [=========>....................] - ETA: 0s - loss: 0.3474 - categorical_accuracy: 0.8994\n",
            "\n",
            "Info su output dell'ultima layer :  KerasTensor(type_spec=TensorSpec(shape=(None, 10), dtype=tf.float32, name=None), name='dense_2/Softmax:0', description=\"created by layer 'dense_2'\")\n",
            "self.params =  {'verbose': 1, 'epochs': 20, 'steps': 120}\n",
            "data.shape :  (60000, 784)\n",
            "\n",
            "\n",
            "120/120 [==============================] - 5s 46ms/step - loss: 0.2969 - categorical_accuracy: 0.9125 - val_loss: 0.1815 - val_categorical_accuracy: 0.9480\n",
            "Epoch 16/20\n",
            "120/120 [==============================] - 1s 11ms/step - loss: 0.2222 - categorical_accuracy: 0.9323 - val_loss: 0.1636 - val_categorical_accuracy: 0.9512\n",
            "Epoch 17/20\n",
            "120/120 [==============================] - 3s 25ms/step - loss: 0.2244 - categorical_accuracy: 0.9302 - val_loss: 0.1535 - val_categorical_accuracy: 0.9562\n",
            "Epoch 18/20\n",
            "  1/120 [..............................] - ETA: 0s - loss: 0.8277 - categorical_accuracy: 0.7500\n",
            "\n",
            "Info su output dell'ultima layer :  KerasTensor(type_spec=TensorSpec(shape=(None, 10), dtype=tf.float32, name=None), name='dense_2/Softmax:0', description=\"created by layer 'dense_2'\")\n",
            "self.params =  {'verbose': 1, 'epochs': 20, 'steps': 120}\n",
            "data.shape :  (60000, 784)\n",
            "\n",
            "\n",
            "\n",
            "\n",
            "Info su output dell'ultima layer :  KerasTensor(type_spec=TensorSpec(shape=(None, 10), dtype=tf.float32, name=None), name='dense_2/Softmax:0', description=\"created by layer 'dense_2'\")\n",
            "self.params =  {'verbose': 1, 'epochs': 20, 'steps': 120}\n",
            "data.shape :  (60000, 784)\n",
            "\n",
            "\n",
            " 31/120 [======>.......................] - ETA: 0s - loss: 0.2393 - categorical_accuracy: 0.9315\n",
            "\n",
            "Info su output dell'ultima layer :  KerasTensor(type_spec=TensorSpec(shape=(None, 10), dtype=tf.float32, name=None), name='dense_2/Softmax:0', description=\"created by layer 'dense_2'\")\n",
            "self.params =  {'verbose': 1, 'epochs': 20, 'steps': 120}\n",
            "data.shape :  (60000, 784)\n",
            "\n",
            "\n",
            "120/120 [==============================] - 2s 14ms/step - loss: 0.2010 - categorical_accuracy: 0.9438 - val_loss: 0.1605 - val_categorical_accuracy: 0.9538\n",
            "Epoch 19/20\n",
            " 37/120 [========>.....................] - ETA: 0s - loss: 0.2024 - categorical_accuracy: 0.9459\n",
            "\n",
            "Info su output dell'ultima layer :  KerasTensor(type_spec=TensorSpec(shape=(None, 10), dtype=tf.float32, name=None), name='dense_2/Softmax:0', description=\"created by layer 'dense_2'\")\n",
            "self.params =  {'verbose': 1, 'epochs': 20, 'steps': 120}\n",
            "data.shape :  (60000, 784)\n",
            "\n",
            "\n",
            " 95/120 [======================>.......] - ETA: 0s - loss: 0.2119 - categorical_accuracy: 0.9395\n",
            "\n",
            "Info su output dell'ultima layer :  KerasTensor(type_spec=TensorSpec(shape=(None, 10), dtype=tf.float32, name=None), name='dense_2/Softmax:0', description=\"created by layer 'dense_2'\")\n",
            "self.params =  {'verbose': 1, 'epochs': 20, 'steps': 120}\n",
            "data.shape :  (60000, 784)\n",
            "\n",
            "\n",
            "110/120 [==========================>...] - ETA: 0s - loss: 0.2103 - categorical_accuracy: 0.9375\n",
            "\n",
            "Info su output dell'ultima layer :  KerasTensor(type_spec=TensorSpec(shape=(None, 10), dtype=tf.float32, name=None), name='dense_2/Softmax:0', description=\"created by layer 'dense_2'\")\n",
            "self.params =  {'verbose': 1, 'epochs': 20, 'steps': 120}\n",
            "data.shape :  (60000, 784)\n",
            "\n",
            "\n",
            "120/120 [==============================] - 2s 14ms/step - loss: 0.2033 - categorical_accuracy: 0.9396 - val_loss: 0.1782 - val_categorical_accuracy: 0.9450\n",
            "Epoch 20/20\n",
            "  7/120 [>.............................] - ETA: 0s - loss: 0.3138 - categorical_accuracy: 0.9286\n",
            "\n",
            "Info su output dell'ultima layer :  KerasTensor(type_spec=TensorSpec(shape=(None, 10), dtype=tf.float32, name=None), name='dense_2/Softmax:0', description=\"created by layer 'dense_2'\")\n",
            "self.params =  {'verbose': 1, 'epochs': 20, 'steps': 120}\n",
            "data.shape :  (60000, 784)\n",
            "\n",
            "\n",
            " 17/120 [===>..........................] - ETA: 0s - loss: 0.3263 - categorical_accuracy: 0.9118\n",
            "\n",
            "Info su output dell'ultima layer :  KerasTensor(type_spec=TensorSpec(shape=(None, 10), dtype=tf.float32, name=None), name='dense_2/Softmax:0', description=\"created by layer 'dense_2'\")\n",
            "self.params =  {'verbose': 1, 'epochs': 20, 'steps': 120}\n",
            "data.shape :  (60000, 784)\n",
            "\n",
            "\n",
            "120/120 [==============================] - 3s 29ms/step - loss: 0.2282 - categorical_accuracy: 0.9260 - val_loss: 0.1512 - val_categorical_accuracy: 0.9547\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Record the training / validation loss / accuracy at each epoch\n",
        "# History attribute containing the lists of successive losses and other metrics.\n",
        "hist = pd.DataFrame(history.history)\n",
        "hist.head(7)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 268
        },
        "id": "4U8RgkwM32VL",
        "outputId": "7b5dd5ef-7527-4139-ed35-2f24a3e3bfa6"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "       loss  categorical_accuracy  val_loss  val_categorical_accuracy\n",
              "0  1.333732              0.597917  0.606044                  0.836833\n",
              "1  0.549126              0.835417  0.403152                  0.884333\n",
              "2  0.470073              0.854167  0.334669                  0.904333\n",
              "3  0.340588              0.904167  0.267819                  0.920333\n",
              "4  0.433520              0.887500  0.302650                  0.912333\n",
              "5  0.332888              0.893750  0.284144                  0.916000\n",
              "6  0.363551              0.893750  0.246698                  0.927333"
            ],
            "text/html": [
              "\n",
              "  <div id=\"df-23873ec9-3eac-43f5-95d7-a6b1368b068b\" class=\"colab-df-container\">\n",
              "    <div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>loss</th>\n",
              "      <th>categorical_accuracy</th>\n",
              "      <th>val_loss</th>\n",
              "      <th>val_categorical_accuracy</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>1.333732</td>\n",
              "      <td>0.597917</td>\n",
              "      <td>0.606044</td>\n",
              "      <td>0.836833</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>0.549126</td>\n",
              "      <td>0.835417</td>\n",
              "      <td>0.403152</td>\n",
              "      <td>0.884333</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>0.470073</td>\n",
              "      <td>0.854167</td>\n",
              "      <td>0.334669</td>\n",
              "      <td>0.904333</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>0.340588</td>\n",
              "      <td>0.904167</td>\n",
              "      <td>0.267819</td>\n",
              "      <td>0.920333</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>0.433520</td>\n",
              "      <td>0.887500</td>\n",
              "      <td>0.302650</td>\n",
              "      <td>0.912333</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>5</th>\n",
              "      <td>0.332888</td>\n",
              "      <td>0.893750</td>\n",
              "      <td>0.284144</td>\n",
              "      <td>0.916000</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>6</th>\n",
              "      <td>0.363551</td>\n",
              "      <td>0.893750</td>\n",
              "      <td>0.246698</td>\n",
              "      <td>0.927333</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>\n",
              "    <div class=\"colab-df-buttons\">\n",
              "\n",
              "  <div class=\"colab-df-container\">\n",
              "    <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-23873ec9-3eac-43f5-95d7-a6b1368b068b')\"\n",
              "            title=\"Convert this dataframe to an interactive table.\"\n",
              "            style=\"display:none;\">\n",
              "\n",
              "  <svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\" viewBox=\"0 -960 960 960\">\n",
              "    <path d=\"M120-120v-720h720v720H120Zm60-500h600v-160H180v160Zm220 220h160v-160H400v160Zm0 220h160v-160H400v160ZM180-400h160v-160H180v160Zm440 0h160v-160H620v160ZM180-180h160v-160H180v160Zm440 0h160v-160H620v160Z\"/>\n",
              "  </svg>\n",
              "    </button>\n",
              "\n",
              "  <style>\n",
              "    .colab-df-container {\n",
              "      display:flex;\n",
              "      gap: 12px;\n",
              "    }\n",
              "\n",
              "    .colab-df-convert {\n",
              "      background-color: #E8F0FE;\n",
              "      border: none;\n",
              "      border-radius: 50%;\n",
              "      cursor: pointer;\n",
              "      display: none;\n",
              "      fill: #1967D2;\n",
              "      height: 32px;\n",
              "      padding: 0 0 0 0;\n",
              "      width: 32px;\n",
              "    }\n",
              "\n",
              "    .colab-df-convert:hover {\n",
              "      background-color: #E2EBFA;\n",
              "      box-shadow: 0px 1px 2px rgba(60, 64, 67, 0.3), 0px 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
              "      fill: #174EA6;\n",
              "    }\n",
              "\n",
              "    .colab-df-buttons div {\n",
              "      margin-bottom: 4px;\n",
              "    }\n",
              "\n",
              "    [theme=dark] .colab-df-convert {\n",
              "      background-color: #3B4455;\n",
              "      fill: #D2E3FC;\n",
              "    }\n",
              "\n",
              "    [theme=dark] .colab-df-convert:hover {\n",
              "      background-color: #434B5C;\n",
              "      box-shadow: 0px 1px 3px 1px rgba(0, 0, 0, 0.15);\n",
              "      filter: drop-shadow(0px 1px 2px rgba(0, 0, 0, 0.3));\n",
              "      fill: #FFFFFF;\n",
              "    }\n",
              "  </style>\n",
              "\n",
              "    <script>\n",
              "      const buttonEl =\n",
              "        document.querySelector('#df-23873ec9-3eac-43f5-95d7-a6b1368b068b button.colab-df-convert');\n",
              "      buttonEl.style.display =\n",
              "        google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
              "\n",
              "      async function convertToInteractive(key) {\n",
              "        const element = document.querySelector('#df-23873ec9-3eac-43f5-95d7-a6b1368b068b');\n",
              "        const dataTable =\n",
              "          await google.colab.kernel.invokeFunction('convertToInteractive',\n",
              "                                                    [key], {});\n",
              "        if (!dataTable) return;\n",
              "\n",
              "        const docLinkHtml = 'Like what you see? Visit the ' +\n",
              "          '<a target=\"_blank\" href=https://colab.research.google.com/notebooks/data_table.ipynb>data table notebook</a>'\n",
              "          + ' to learn more about interactive tables.';\n",
              "        element.innerHTML = '';\n",
              "        dataTable['output_type'] = 'display_data';\n",
              "        await google.colab.output.renderOutput(dataTable, element);\n",
              "        const docLink = document.createElement('div');\n",
              "        docLink.innerHTML = docLinkHtml;\n",
              "        element.appendChild(docLink);\n",
              "      }\n",
              "    </script>\n",
              "  </div>\n",
              "\n",
              "\n",
              "<div id=\"df-8bb8e66c-439c-42ae-b436-29e832a55346\">\n",
              "  <button class=\"colab-df-quickchart\" onclick=\"quickchart('df-8bb8e66c-439c-42ae-b436-29e832a55346')\"\n",
              "            title=\"Suggest charts\"\n",
              "            style=\"display:none;\">\n",
              "\n",
              "<svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\"viewBox=\"0 0 24 24\"\n",
              "     width=\"24px\">\n",
              "    <g>\n",
              "        <path d=\"M19 3H5c-1.1 0-2 .9-2 2v14c0 1.1.9 2 2 2h14c1.1 0 2-.9 2-2V5c0-1.1-.9-2-2-2zM9 17H7v-7h2v7zm4 0h-2V7h2v10zm4 0h-2v-4h2v4z\"/>\n",
              "    </g>\n",
              "</svg>\n",
              "  </button>\n",
              "\n",
              "<style>\n",
              "  .colab-df-quickchart {\n",
              "      --bg-color: #E8F0FE;\n",
              "      --fill-color: #1967D2;\n",
              "      --hover-bg-color: #E2EBFA;\n",
              "      --hover-fill-color: #174EA6;\n",
              "      --disabled-fill-color: #AAA;\n",
              "      --disabled-bg-color: #DDD;\n",
              "  }\n",
              "\n",
              "  [theme=dark] .colab-df-quickchart {\n",
              "      --bg-color: #3B4455;\n",
              "      --fill-color: #D2E3FC;\n",
              "      --hover-bg-color: #434B5C;\n",
              "      --hover-fill-color: #FFFFFF;\n",
              "      --disabled-bg-color: #3B4455;\n",
              "      --disabled-fill-color: #666;\n",
              "  }\n",
              "\n",
              "  .colab-df-quickchart {\n",
              "    background-color: var(--bg-color);\n",
              "    border: none;\n",
              "    border-radius: 50%;\n",
              "    cursor: pointer;\n",
              "    display: none;\n",
              "    fill: var(--fill-color);\n",
              "    height: 32px;\n",
              "    padding: 0;\n",
              "    width: 32px;\n",
              "  }\n",
              "\n",
              "  .colab-df-quickchart:hover {\n",
              "    background-color: var(--hover-bg-color);\n",
              "    box-shadow: 0 1px 2px rgba(60, 64, 67, 0.3), 0 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
              "    fill: var(--button-hover-fill-color);\n",
              "  }\n",
              "\n",
              "  .colab-df-quickchart-complete:disabled,\n",
              "  .colab-df-quickchart-complete:disabled:hover {\n",
              "    background-color: var(--disabled-bg-color);\n",
              "    fill: var(--disabled-fill-color);\n",
              "    box-shadow: none;\n",
              "  }\n",
              "\n",
              "  .colab-df-spinner {\n",
              "    border: 2px solid var(--fill-color);\n",
              "    border-color: transparent;\n",
              "    border-bottom-color: var(--fill-color);\n",
              "    animation:\n",
              "      spin 1s steps(1) infinite;\n",
              "  }\n",
              "\n",
              "  @keyframes spin {\n",
              "    0% {\n",
              "      border-color: transparent;\n",
              "      border-bottom-color: var(--fill-color);\n",
              "      border-left-color: var(--fill-color);\n",
              "    }\n",
              "    20% {\n",
              "      border-color: transparent;\n",
              "      border-left-color: var(--fill-color);\n",
              "      border-top-color: var(--fill-color);\n",
              "    }\n",
              "    30% {\n",
              "      border-color: transparent;\n",
              "      border-left-color: var(--fill-color);\n",
              "      border-top-color: var(--fill-color);\n",
              "      border-right-color: var(--fill-color);\n",
              "    }\n",
              "    40% {\n",
              "      border-color: transparent;\n",
              "      border-right-color: var(--fill-color);\n",
              "      border-top-color: var(--fill-color);\n",
              "    }\n",
              "    60% {\n",
              "      border-color: transparent;\n",
              "      border-right-color: var(--fill-color);\n",
              "    }\n",
              "    80% {\n",
              "      border-color: transparent;\n",
              "      border-right-color: var(--fill-color);\n",
              "      border-bottom-color: var(--fill-color);\n",
              "    }\n",
              "    90% {\n",
              "      border-color: transparent;\n",
              "      border-bottom-color: var(--fill-color);\n",
              "    }\n",
              "  }\n",
              "</style>\n",
              "\n",
              "  <script>\n",
              "    async function quickchart(key) {\n",
              "      const quickchartButtonEl =\n",
              "        document.querySelector('#' + key + ' button');\n",
              "      quickchartButtonEl.disabled = true;  // To prevent multiple clicks.\n",
              "      quickchartButtonEl.classList.add('colab-df-spinner');\n",
              "      try {\n",
              "        const charts = await google.colab.kernel.invokeFunction(\n",
              "            'suggestCharts', [key], {});\n",
              "      } catch (error) {\n",
              "        console.error('Error during call to suggestCharts:', error);\n",
              "      }\n",
              "      quickchartButtonEl.classList.remove('colab-df-spinner');\n",
              "      quickchartButtonEl.classList.add('colab-df-quickchart-complete');\n",
              "    }\n",
              "    (() => {\n",
              "      let quickchartButtonEl =\n",
              "        document.querySelector('#df-8bb8e66c-439c-42ae-b436-29e832a55346 button');\n",
              "      quickchartButtonEl.style.display =\n",
              "        google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
              "    })();\n",
              "  </script>\n",
              "</div>\n",
              "    </div>\n",
              "  </div>\n"
            ],
            "application/vnd.google.colaboratory.intrinsic+json": {
              "type": "dataframe",
              "variable_name": "hist",
              "summary": "{\n  \"name\": \"hist\",\n  \"rows\": 20,\n  \"fields\": [\n    {\n      \"column\": \"loss\",\n      \"properties\": {\n        \"dtype\": \"number\",\n        \"std\": 0.24915138148926083,\n        \"min\": 0.20104044675827026,\n        \"max\": 1.3337324857711792,\n        \"num_unique_values\": 20,\n        \"samples\": [\n          1.3337324857711792,\n          0.20104044675827026,\n          0.22218482196331024\n        ],\n        \"semantic_type\": \"\",\n        \"description\": \"\"\n      }\n    },\n    {\n      \"column\": \"categorical_accuracy\",\n      \"properties\": {\n        \"dtype\": \"number\",\n        \"std\": 0.07479174149191287,\n        \"min\": 0.5979166626930237,\n        \"max\": 0.9437500238418579,\n        \"num_unique_values\": 18,\n        \"samples\": [\n          0.5979166626930237,\n          0.8354166746139526,\n          0.9291666746139526\n        ],\n        \"semantic_type\": \"\",\n        \"description\": \"\"\n      }\n    },\n    {\n      \"column\": \"val_loss\",\n      \"properties\": {\n        \"dtype\": \"number\",\n        \"std\": 0.10837436935344083,\n        \"min\": 0.1512046754360199,\n        \"max\": 0.6060442328453064,\n        \"num_unique_values\": 20,\n        \"samples\": [\n          0.6060442328453064,\n          0.16049394011497498,\n          0.16363099217414856\n        ],\n        \"semantic_type\": \"\",\n        \"description\": \"\"\n      }\n    },\n    {\n      \"column\": \"val_categorical_accuracy\",\n      \"properties\": {\n        \"dtype\": \"number\",\n        \"std\": 0.0290754137345343,\n        \"min\": 0.8368333578109741,\n        \"max\": 0.956166684627533,\n        \"num_unique_values\": 19,\n        \"samples\": [\n          0.8368333578109741,\n          0.9160000085830688,\n          0.9418333172798157\n        ],\n        \"semantic_type\": \"\",\n        \"description\": \"\"\n      }\n    }\n  ]\n}"
            }
          },
          "metadata": {},
          "execution_count": 10
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Conserva i valori dei model's weight e bias.\n",
        "V = model.get_weights()"
      ],
      "metadata": {
        "id": "JMvvDS634dHL"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "Salva tutto modello su file"
      ],
      "metadata": {
        "id": "S7whoeq4tdiP"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "#Memorizza la rete neurale completa nel formato file HDF5\n",
        "save_path = \".\"\n",
        "model.save(os.path.join(save_path,\"network.h5\"))"
      ],
      "metadata": {
        "id": "6LYDi0uptlB7",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "a6309150-c477-440c-ad89-b83cfd61244e"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "/usr/local/lib/python3.10/dist-packages/keras/src/engine/training.py:3103: UserWarning: You are saving your model as an HDF5 file via `model.save()`. This file format is considered legacy. We recommend using instead the native Keras format, e.g. `model.save('my_model.keras')`.\n",
            "  saving_api.save_model(\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "####ModelCheckpoint\n",
        "\n",
        "https://keras.io/api/callbacks/model_checkpoint/"
      ],
      "metadata": {
        "id": "90Ciuc_oGoPC"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "checkpoint_filepath = '/tmp/ckpt/checkpoint.model.keras'\n",
        "model_checkpoint_callback = tf.keras.callbacks.ModelCheckpoint(\n",
        "    filepath=checkpoint_filepath,\n",
        "    monitor='val_loss',\n",
        "    mode='min',\n",
        "    save_best_only=True)\n",
        "\n",
        "# Model is saved at the end of every epoch, if it's the best seen so far.\n",
        "num_iterations = 30\n",
        "my_batch_size = 64\n",
        "\n",
        "model.fit(x=xtrain,\n",
        "             y=label_train,\n",
        "             batch_size=my_batch_size,\n",
        "             epochs=num_iterations,\n",
        "             shuffle=True,\n",
        "             steps_per_epoch = None,\n",
        "             validation_split=0.1,\n",
        "             verbose=0,\n",
        "             callbacks=[model_checkpoint_callback]) # # Pass callback to training\n",
        "\n",
        "# The model (that are considered the best) can be loaded as -\n",
        "my2model = tf.keras.models.load_model(checkpoint_filepath)"
      ],
      "metadata": {
        "id": "XgqMBVWMG8qB"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# Alternatively, one could checkpoint just the model weights as -\n",
        "checkpoint_filepath = '/tmp/ckpt/checkpoint.weights.h5'\n",
        "model_checkpoint_callback = tf.keras.callbacks.ModelCheckpoint(\n",
        "    filepath=checkpoint_filepath,\n",
        "    save_weights_only=True,\n",
        "    monitor='val_accuracy',\n",
        "    mode='max',\n",
        "    save_best_only=True)\n",
        "\n",
        "# Model weights are saved at the end of every epoch, if it's the best seen\n",
        "# so far.\n",
        "num_iterations = 30\n",
        "my_batch_size = 64\n",
        "\n",
        "model.fit(x=xtrain,\n",
        "             y=label_train,\n",
        "             batch_size=my_batch_size,\n",
        "             epochs=num_iterations,\n",
        "             shuffle=True,\n",
        "             steps_per_epoch = None,\n",
        "             validation_split=0.1,\n",
        "             verbose=0,\n",
        "             callbacks=[model_checkpoint_callback])\n",
        "\n",
        "# The model weights (that are considered the best) can be loaded as -\n",
        "my2model = model.load_weights(checkpoint_filepath)"
      ],
      "metadata": {
        "id": "owG5idHMI1VC"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "####Carica il modello salvato"
      ],
      "metadata": {
        "id": "G9yLZ5SQ6FHp"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "NN2 = tf.keras.models.load_model(os.path.join(save_path,\"network.h5\"))"
      ],
      "metadata": {
        "id": "zgKL0S-w6IYy"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "#Test se i modelli coincidono\n",
        "ri = np.random.randint(low=0, high=len(xtest))\n",
        "\n",
        "x = xtest[ri]\n",
        "x = tf.reshape(x, [1, -1])\n",
        "\n",
        "pred1 =  model.predict(x)\n",
        "pred2 = NN2.predict(x)\n",
        "\n",
        "print(pred1 == pred2)"
      ],
      "metadata": {
        "id": "YQ23g1zXthDB",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "9f04f982-f763-4b8a-c224-d6bce9f6ca5e"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "1/1 [==============================] - 0s 375ms/step\n",
            "1/1 [==============================] - 0s 454ms/step\n",
            "[[ True  True  True  True  True  True  True  True  True  True]]\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "####Test Campione"
      ],
      "metadata": {
        "id": "Nqj_P_9I5ZZr"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "ri = np.random.randint(low=0, high=len(xtest))\n",
        "\n",
        "x = xtest[ri]\n",
        "\n",
        "xf = x.reshape(28, 28)\n",
        "plt.figure(figsize=[0.7, 0.7])\n",
        "plt.imshow(xf, cmap=plt.get_cmap('gray'))\n",
        "plt.axis('off')\n",
        "plt.show()\n",
        "\n",
        "x = tf.reshape(x, [1, -1])\n",
        "y = label_test[ri]\n",
        "yy = model.predict(x)\n",
        "print(\"TrueLabel:  \",y)\n",
        "print(\"Previsione: \",yy.round(2))\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 144
        },
        "id": "OXhUzktDy7Bc",
        "outputId": "0ddf23fd-7721-4514-8847-9087e33d1876"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 70x70 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "1/1 [==============================] - 0s 27ms/step\n",
            "TrueLabel:   [0. 0. 0. 0. 0. 0. 0. 0. 1. 0.]\n",
            "Previsione:  [[0.   0.   0.   0.   0.   0.01 0.   0.   0.99 0.  ]]\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "####Accuracy"
      ],
      "metadata": {
        "id": "AKr5Jcv5LJLo"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "score = model.evaluate(xtest, label_test, len(xtest)) # calcola accuracy di tutto il dataset test in una volta sola - sfruttare la Vettorizzazione\n",
        "# model.evaluate(xtest, label_test,1) # calcola accuracy di un elemento per volta di tutto il dataset test, stesso risultato di sopra ma necessita più tempo\n",
        "\n",
        "print(\"Test loss:\", score[0])\n",
        "print(\"Test accuracy:\", score[1])\n"
      ],
      "metadata": {
        "id": "_0ZmWO1V4XRh",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "235829e2-cebb-4f8f-d50a-39f9ed8bc28f"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "1/1 [==============================] - 0s 256ms/step - loss: 0.1850 - categorical_accuracy: 0.9443\n",
            "Test loss: 0.1849636435508728\n",
            "Test accuracy: 0.9442999958992004\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# tf.keras.metrics.Accuracy() - vs. -  tf.keras.metrics.CategoricalAccuracy()\n",
        "\n",
        "m1 = tf.keras.metrics.Accuracy()\n",
        "m1.update_state(model.predict(xtest), label_test)\n",
        "a  = m1.result()\n",
        "print(\"Accuracy su tutto il test dataset :\",a.numpy())\n",
        "\n",
        "print(\"\")\n",
        "\n",
        "m2 = tf.keras.metrics.CategoricalAccuracy()\n",
        "\n",
        "m2.update_state(model.predict(xtest), label_test)\n",
        "a2  = m2.result()\n",
        "print(\"Accuracy su tutto il test dataset :\",a2.numpy())\n",
        "\n"
      ],
      "metadata": {
        "id": "zj5nFmwRK8aP",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "1b4995d7-b064-43a0-f952-4e75a025978b"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "313/313 [==============================] - 2s 6ms/step\n",
            "Accuracy su tutto il test dataset : 0.0\n",
            "\n",
            "313/313 [==============================] - 1s 3ms/step\n",
            "Accuracy su tutto il test dataset : 0.9443\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "tf.keras.metrics.CategoricalAccuracy() offre lo stesso risultato del codice sottostante, con un notevole risparmio di tempo.\n",
        "\n",
        "Mentre tf.keras.metrics.Accuracy() richiede che le predizioni coincidano perfettamente con le etichette reali per considerarle corrette (necessita di una corrispondenza esatta!)"
      ],
      "metadata": {
        "id": "TycFbnRdWQTZ"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "# codice di sotto, fa lo stesso lavoro di tf.keras.metrics.CategoricalAccuracy()\n",
        "c=0\n",
        "lt =  len(xtest)\n",
        "\n",
        "for i in range(lt):\n",
        "  x = tf.reshape(xtest[i], [1, -1])\n",
        "  predicted_digit = tf.argmax(my_model.predict(x),1)\n",
        "  actual_digit = tf.argmax(label_test[i])\n",
        "  if predicted_digit.numpy()[0] == actual_digit.numpy():\n",
        "     c = c + 1\n",
        "print(\"Accuracy su tutto il test dataset :\", c/lt)"
      ],
      "metadata": {
        "id": "8JSrFXt-EAjB"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "####Grafici"
      ],
      "metadata": {
        "id": "g3t-SwNE5b__"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "# list all data in history\n",
        "print(history.history.keys())"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "GGCO7kCAumqe",
        "outputId": "a8b7a637-fe63-42ee-9ab8-432585d14a8a"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "dict_keys(['loss', 'categorical_accuracy', 'val_loss', 'val_categorical_accuracy'])\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "plt.figure(figsize=(3, 3))\n",
        "\n",
        "plt.title('Summary of loss')\n",
        "plt.ylabel('loss value')\n",
        "plt.xlabel('epoch')\n",
        "\n",
        "plt.plot(range(num_iterations), history.history['val_loss'])"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 351
        },
        "id": "EPuVF6Wqux0P",
        "outputId": "60d0582f-27ff-49de-a55a-5e621129d8e1"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[<matplotlib.lines.Line2D at 0x7d1e016abc70>]"
            ]
          },
          "metadata": {},
          "execution_count": 27
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 300x300 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "plt.figure(figsize=(3.4, 3.4))\n",
        "\n",
        "plt.title('Summary accuracy')\n",
        "plt.ylabel('categorical_accuracy')\n",
        "plt.xlabel('epoch')\n",
        "\n",
        "\n",
        "plt.plot(history.history['categorical_accuracy'], label='Train_acc', color='green')\n",
        "plt.plot(history.history['val_categorical_accuracy'],label='Valid_Train_acc', color='orange')\n",
        "\n",
        "plt.legend(bbox_to_anchor=(1.05, 1), loc='upper left', borderaxespad=0, title='Metric' )\n",
        "plt.show()\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 364
        },
        "id": "rlsQFlH_1DD7",
        "outputId": "f2c6fa37-aa32-4f9e-b2f9-fa76e6c515a8"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 340x340 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "####Note conclusive\n",
        "\n",
        "\n",
        "\n",
        "In questa sezione non ho trattato la **Regularization**, un insieme di tecniche volte a ridurre l'overfitting nei modelli di machine learning.\n",
        "\n",
        "La natura introduttiva del corso ha reso opportuno non approfondire questo argomento, che comunque merita attenzione. In generale, le tecniche di Regularization riducono leggermente l'accuratezza del modello durante l'addestramento, ma ne migliorano la capacità di generalizzare a nuovi dati. Esistono due categorie principali; norm-based: (esempio: L1 o L2 Regularization) utilizza la funzione di loss per penalizzare modelli.\n",
        "Stochastic: questa categoria include tecniche come il Dropout, che disattivano casualmente alcuni neuroni durante l'addestramento."
      ],
      "metadata": {
        "id": "tKodrw1hZsLo"
      }
    }
  ]
}