4-CNN.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Convolutional Neural Networks"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## A theoretical overview\n",
    "Convolutional neural networks are specifically designed to analyze images with layers developed to emulate the functioning of the visual cortex in biological organisms. They analyze separately different regions of the image in order to recognize patterns indipendently from their position and reduce the overall complexity of the model to avoid overfitting.\n",
    "\n",
    "The architecture of the models used in this notebook are essentially analogous to a feedforward neural network with the addition of two new types of layer: convolutional layers and pooling layers. These layers are usually used in blocks of some convolutional layers followed by a pooling one, gradually reducing the dimensions of the features' maps and incrementing the number of filters (see some examples of established architectures in Géron, 2019). Finally all the maps are flattened and connected to dense layers which reach the output layer of the necessary size.\n",
    "\n",
    "### Convolutional layer\n",
    "In a 2D convolutional layer the parameters are organized in a set of filters, which are matrices of smaller dimensions than the input matrix. The weights contained in these matrices get updated during the training process. The layer works as follows: each filter passes along the input matrix generating a feature map in which each pixel's value is obtained from the convolution of the filter with the region of the input matrix considered at the respective step. The convolution of the two matrix, of the same dimensions, consists in an element-wise multiplication followed by the summation of all the obtained values.\n",
    "This process is shown in the following image (from _Machine learning glossary_, n.d.).\n",
    "\n",
    "<center>\n",
    "    <br/>\n",
    "    <img src=\"./images/AnimatedConvolution.gif\" width=\"30%\">\n",
    "</center>\n",
    "\n",
    "After the convolution an activation function could also be applied.\n",
    "\n",
    "### Pooling layers \n",
    "Pooling layers behave in the same way of convolutional layers, i.e. considering step by step a different slice of the input matrix, but do not have any trainable parameter.\n",
    "Instead, the feature map is generated taking the maximum pixel's value in the region considered (max pooling layers) or averaging those values (average pooling layers).\n",
    "The aim of these layers is usually that of reducing the size of the feature maps trying to condensate the information stored in the image. For example, neighboring pixels are often highly correlated assuming very similar values, a pooling layers helps removing the redondant data.\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Hyperparameters\n",
    "\n",
    "Most hyperparameters of convolutional neural networks are analogous to the ones of feed forward networks. They include the activation functions, the optimizer with its hyperparameters and the choice of the overall architecture requiring to choose the number of layers, their dimensions and the order in which they appear.\n",
    "\n",
    "As in the previous notebook, for feedforward neural networks, I have used the Adam optimizer, with the default values of its hyperparameters, which seemed to work well. Some tests made with different optimizers have not shown better performances. I also early stopped the training, monitoring the validation loss and interrupting the process if it does not improve for 7 epochs.\n",
    "\n",
    "The new hyperparameters regards convolutional layers. In addition to the filters' size and the number of feature's map it is necessary to define strides and padding. The former defines the number of pixels the filter is moved vertically and horizontally during its application to the input. The latter defines the number of padding pixels to add to the input matrix which can be required to make sure that the filter does not fall out of the input image.\n",
    "In the models trained in this notebook I have used the `padding=same` parameter which automatically sets the padding to produce feature maps of the same dimensions of the input (see _Conv2D layer_, n.d.).\n",
    "\n",
    "## Trained and tested models\n",
    "\n",
    "After some tests aimed at exploring how different choices lead to different performances, I have decided to present and compare in this notebook the following three different models.\n",
    "\n",
    "### LeNet5\n",
    "LeNet5 is one of the earliest convolutional neural networks published, firstly applied to handwritten zipe code recognition (LeCun, 1989). I have kept the same structure of the original model keeping the same filter's size. However,since the input images have different dimensions, the features maps result in different sizes too.\n",
    "Like the original paper I have used the `tanh()` activation functions in the convolutional layers, and average pooling layers instead of today's more common max pooling layers.\n",
    "I have however modified the activation function of the output layer using `softmax` like in all the other models.\n",
    "\n",
    "The model is summarized in this image\n",
    "<center>\n",
    "    <br/>\n",
    "    <img src=\"./images/leNet.png\" width=\"70%\">\n",
    "</center>\n",
    "\n",
    "### singleConvCNN\n",
    "This neural network it is not instead an established model. However in its design I took inspiration from the general features shared by most of the well known architectures (Géron, 2019).\n",
    "It is composed of three blocks of a convolutional layer followed by a max pooling layer. Each max pooling layer reduces by half both the dimensions of the input image while the convolutional layer of the following block doubles the depth, i.e. the number of features maps. \n",
    "After three of these blocks the data tensor is flattened and connected to two dense layers gradually reducing their dimensions up to that of the output layer which is always a 22 neurons, softmax, dense layer.\n",
    "\n",
    "The network, summarized above, is shown in this image\n",
    "<center>\n",
    "    <br/>\n",
    "    <img src=\"./images/singleCNN.png\" width=\"70%\">\n",
    "</center>\n",
    "\n",
    "### doubleConvCNN \n",
    "This last model is very similar to the precedent with a simple modification that increments its complexity:\n",
    "each block contains two identical convolution layers before of the respective max pooling layer.\n",
    "\n",
    "A summary of this network is shown in this image:\n",
    "<center>\n",
    "    <br/>\n",
    "    <img src=\"./images/doubleCNN.png\" width=\"100%\">\n",
    "</center>\n",
    "\n",
    "\n",
    "I have also tried to train the same models with the addition of a dropout (with rate=0.5) layer after the flattening of the data and before the part of the network with dense layers. \n",
    "The dropout layer randomly drops, during the training process, a fraction (rate) of the connections between the two layers where is placed. The connections to drop are randomly selected at each step of the training process. When the neural network is trained it gets deployed with all the connections restored.\n",
    "This should force the neural network to not rely too much on single connections and help thus producing more robust models resulting in lower generalization errors.\n",
    "\n",
    "All these models are defined in `models.py`. In the following cell they are loaded and their summary generated by tensorflow is printed showing also the number of trainable parameters of each of them.\n",
    "Compared to all the feed forward models used in this project, these convolutional networks are more complex with a larger number of trainable parameters. However, in this case all the 71x60x3 image is provided in input to the network without further reduction."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "# imports\n",
    "import my_utils as u\n",
    "import numpy as np\n",
    "import tensorflow as tf"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"singleConvCNN\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "input_5 (InputLayer)         [(None, 60, 71, 3)]       0         \n",
      "_________________________________________________________________\n",
      "b1_c1 (Conv2D)               (None, 60, 71, 8)         224       \n",
      "_________________________________________________________________\n",
      "b1_p (MaxPooling2D)          (None, 30, 36, 8)         0         \n",
      "_________________________________________________________________\n",
      "b2_c1 (Conv2D)               (None, 30, 36, 16)        1168      \n",
      "_________________________________________________________________\n",
      "b2_p (MaxPooling2D)          (None, 15, 18, 16)        0         \n",
      "_________________________________________________________________\n",
      "b3_c1 (Conv2D)               (None, 15, 18, 32)        4640      \n",
      "_________________________________________________________________\n",
      "b3_p (MaxPooling2D)          (None, 8, 9, 32)          0         \n",
      "_________________________________________________________________\n",
      "flatten (Flatten)            (None, 2304)              0         \n",
      "_________________________________________________________________\n",
      "FC1 (Dense)                  (None, 128)               295040    \n",
      "_________________________________________________________________\n",
      "FC2 (Dense)                  (None, 64)                8256      \n",
      "_________________________________________________________________\n",
      "Out (Dense)                  (None, 22)                1430      \n",
      "=================================================================\n",
      "Total params: 310,758\n",
      "Trainable params: 310,758\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "Model: \"doubleConvCNN\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "input_6 (InputLayer)         [(None, 60, 71, 3)]       0         \n",
      "_________________________________________________________________\n",
      "b1_c1 (Conv2D)               (None, 60, 71, 8)         224       \n",
      "_________________________________________________________________\n",
      "b1_c2 (Conv2D)               (None, 60, 71, 8)         584       \n",
      "_________________________________________________________________\n",
      "b1_p (MaxPooling2D)          (None, 30, 35, 8)         0         \n",
      "_________________________________________________________________\n",
      "b2_c1 (Conv2D)               (None, 30, 35, 16)        1168      \n",
      "_________________________________________________________________\n",
      "b2_c2 (Conv2D)               (None, 30, 35, 16)        2320      \n",
      "_________________________________________________________________\n",
      "b2_p (MaxPooling2D)          (None, 15, 17, 16)        0         \n",
      "_________________________________________________________________\n",
      "b3_c1 (Conv2D)               (None, 15, 17, 32)        4640      \n",
      "_________________________________________________________________\n",
      "b3_c2 (Conv2D)               (None, 15, 17, 32)        9248      \n",
      "_________________________________________________________________\n",
      "b3_p (MaxPooling2D)          (None, 7, 8, 32)          0         \n",
      "_________________________________________________________________\n",
      "flatten (Flatten)            (None, 1792)              0         \n",
      "_________________________________________________________________\n",
      "FC1 (Dense)                  (None, 128)               229504    \n",
      "_________________________________________________________________\n",
      "FC2 (Dense)                  (None, 64)                8256      \n",
      "_________________________________________________________________\n",
      "Out (Dense)                  (None, 22)                1430      \n",
      "=================================================================\n",
      "Total params: 257,374\n",
      "Trainable params: 257,374\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "Model: \"LeNet5\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "In (InputLayer)              [(None, 60, 71, 3)]       0         \n",
      "_________________________________________________________________\n",
      "C1 (Conv2D)                  (None, 56, 67, 6)         456       \n",
      "_________________________________________________________________\n",
      "S2 (AveragePooling2D)        (None, 28, 33, 6)         0         \n",
      "_________________________________________________________________\n",
      "C3 (Conv2D)                  (None, 24, 29, 16)        2416      \n",
      "_________________________________________________________________\n",
      "S4 (AveragePooling2D)        (None, 12, 14, 16)        0         \n",
      "_________________________________________________________________\n",
      "C5 (Conv2D)                  (None, 12, 14, 120)       2040      \n",
      "_________________________________________________________________\n",
      "flatten_2 (Flatten)          (None, 20160)             0         \n",
      "_________________________________________________________________\n",
      "F6 (Dense)                   (None, 84)                1693524   \n",
      "_________________________________________________________________\n",
      "Out (Dense)                  (None, 22)                1870      \n",
      "=================================================================\n",
      "Total params: 1,700,306\n",
      "Trainable params: 1,700,306\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "Model: \"singleConvDroputCNN\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "input_7 (InputLayer)         [(None, 60, 71, 3)]       0         \n",
      "_________________________________________________________________\n",
      "b1_c1 (Conv2D)               (None, 60, 71, 8)         224       \n",
      "_________________________________________________________________\n",
      "b1_p (MaxPooling2D)          (None, 30, 36, 8)         0         \n",
      "_________________________________________________________________\n",
      "b2_c1 (Conv2D)               (None, 30, 36, 16)        1168      \n",
      "_________________________________________________________________\n",
      "b2_p (MaxPooling2D)          (None, 15, 18, 16)        0         \n",
      "_________________________________________________________________\n",
      "b3_c1 (Conv2D)               (None, 15, 18, 32)        4640      \n",
      "_________________________________________________________________\n",
      "b3_p (MaxPooling2D)          (None, 8, 9, 32)          0         \n",
      "_________________________________________________________________\n",
      "flatten (Flatten)            (None, 2304)              0         \n",
      "_________________________________________________________________\n",
      "dropout_3 (Dropout)          (None, 2304)              0         \n",
      "_________________________________________________________________\n",
      "FC1 (Dense)                  (None, 128)               295040    \n",
      "_________________________________________________________________\n",
      "FC2 (Dense)                  (None, 64)                8256      \n",
      "_________________________________________________________________\n",
      "Out (Dense)                  (None, 22)                1430      \n",
      "=================================================================\n",
      "Total params: 310,758\n",
      "Trainable params: 310,758\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "Model: \"doubleConvDroputCNN\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "input_8 (InputLayer)         [(None, 60, 71, 3)]       0         \n",
      "_________________________________________________________________\n",
      "b1_c1 (Conv2D)               (None, 60, 71, 8)         224       \n",
      "_________________________________________________________________\n",
      "b1_c2 (Conv2D)               (None, 60, 71, 8)         584       \n",
      "_________________________________________________________________\n",
      "b1_p (MaxPooling2D)          (None, 30, 35, 8)         0         \n",
      "_________________________________________________________________\n",
      "b2_c1 (Conv2D)               (None, 30, 35, 16)        1168      \n",
      "_________________________________________________________________\n",
      "b2_c2 (Conv2D)               (None, 30, 35, 16)        2320      \n",
      "_________________________________________________________________\n",
      "b2_p (MaxPooling2D)          (None, 15, 17, 16)        0         \n",
      "_________________________________________________________________\n",
      "b3_c1 (Conv2D)               (None, 15, 17, 32)        4640      \n",
      "_________________________________________________________________\n",
      "b3_c2 (Conv2D)               (None, 15, 17, 32)        9248      \n",
      "_________________________________________________________________\n",
      "b3_p (MaxPooling2D)          (None, 7, 8, 32)          0         \n",
      "_________________________________________________________________\n",
      "flatten (Flatten)            (None, 1792)              0         \n",
      "_________________________________________________________________\n",
      "dropout_4 (Dropout)          (None, 1792)              0         \n",
      "_________________________________________________________________\n",
      "FC1 (Dense)                  (None, 128)               229504    \n",
      "_________________________________________________________________\n",
      "FC2 (Dense)                  (None, 64)                8256      \n",
      "_________________________________________________________________\n",
      "Out (Dense)                  (None, 22)                1430      \n",
      "=================================================================\n",
      "Total params: 257,374\n",
      "Trainable params: 257,374\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "Model: \"LeNet5_Dropout\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "In (InputLayer)              [(None, 60, 71, 3)]       0         \n",
      "_________________________________________________________________\n",
      "C1 (Conv2D)                  (None, 56, 67, 6)         456       \n",
      "_________________________________________________________________\n",
      "S2 (AveragePooling2D)        (None, 28, 33, 6)         0         \n",
      "_________________________________________________________________\n",
      "C3 (Conv2D)                  (None, 24, 29, 16)        2416      \n",
      "_________________________________________________________________\n",
      "S4 (AveragePooling2D)        (None, 12, 14, 16)        0         \n",
      "_________________________________________________________________\n",
      "C5 (Conv2D)                  (None, 12, 14, 120)       2040      \n",
      "_________________________________________________________________\n",
      "flatten_3 (Flatten)          (None, 20160)             0         \n",
      "_________________________________________________________________\n",
      "dropout_5 (Dropout)          (None, 20160)             0         \n",
      "_________________________________________________________________\n",
      "F6 (Dense)                   (None, 84)                1693524   \n",
      "_________________________________________________________________\n",
      "Out (Dense)                  (None, 22)                1870      \n",
      "=================================================================\n",
      "Total params: 1,700,306\n",
      "Trainable params: 1,700,306\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "import models\n",
    "\n",
    "models_totry = [models.singleConvCNN(), models.doubleConvCNN(), models.LeNet5(), models.singleConvCNN(dropout=True), models.doubleConvCNN(dropout=True), models.LeNet5(dropout=True)]\n",
    "for model in models_totry:\n",
    "    model.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Training"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████████████████████████████████| 22348/22348 [00:23<00:00, 950.14it/s]\n"
     ]
    }
   ],
   "source": [
    "#loading training data\n",
    "x_train, y_train, df_data = u.get_images(u.train_files)\n",
    "#rescaling pixel values in the [0;1] range\n",
    "x_train = x_train/256"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|████████████████████████████████████████| 4720/4720 [00:04<00:00, 992.67it/s]\n"
     ]
    }
   ],
   "source": [
    "#loading test data\n",
    "x_test, y_test, df_data = u.get_images(u.test_files)\n",
    "#rescaling pixel values in the [0;1] range\n",
    "x_test = x_test/256"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(22348, 60, 71, 3)"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "x_train.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "#changing format of labels\n",
    "y_train_mrx = np.array([[int(i==y) for i in range(0,22)] for y in y_train])\n",
    "y_test_mrx = np.array([[int(i==y) for i in range(0,22)] for y in y_test])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "tags": [
     "outputPrepend"
    ]
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " 0.2825 - val_loss: 0.3002\n",
      "Epoch 3/20\n",
      "699/699 [==============================] - 68s 97ms/step - loss: 0.2084 - val_loss: 0.2370\n",
      "Epoch 4/20\n",
      "699/699 [==============================] - 68s 97ms/step - loss: 0.1622 - val_loss: 0.2570\n",
      "Epoch 5/20\n",
      "699/699 [==============================] - 68s 97ms/step - loss: 0.1474 - val_loss: 0.2467\n",
      "Epoch 6/20\n",
      "699/699 [==============================] - 69s 98ms/step - loss: 0.1187 - val_loss: 0.2248\n",
      "Epoch 7/20\n",
      "699/699 [==============================] - 71s 101ms/step - loss: 0.1062 - val_loss: 0.2492\n",
      "Epoch 8/20\n",
      "699/699 [==============================] - 71s 101ms/step - loss: 0.0925 - val_loss: 0.2693\n",
      "Epoch 9/20\n",
      "699/699 [==============================] - 71s 102ms/step - loss: 0.0857 - val_loss: 0.3185\n",
      "Epoch 10/20\n",
      "699/699 [==============================] - 74s 106ms/step - loss: 0.0713 - val_loss: 0.3286\n",
      "Epoch 11/20\n",
      "699/699 [==============================] - 72s 103ms/step - loss: 0.0741 - val_loss: 0.3413\n",
      "Epoch 12/20\n",
      "699/699 [==============================] - 73s 104ms/step - loss: 0.0617 - val_loss: 0.3578\n",
      "Epoch 13/20\n",
      "699/699 [==============================] - 73s 104ms/step - loss: 0.0529 - val_loss: 0.3367\n",
      "Epoch 00013: early stopping\n",
      "fit completed, saving model\n",
      "2022-02-01 23:32:06.698033: W tensorflow/python/util/util.cc:348] Sets are not currently considered sequences, but this may change in the future, so consider avoiding using them.\n",
      "INFO:tensorflow:Assets written to: singleConvCNN.CNN/assets\n",
      "Model doubleConvCNN prepared. Architecture:\n",
      "Model: \"doubleConvCNN\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "input_2 (InputLayer)         [(None, 60, 71, 3)]       0         \n",
      "_________________________________________________________________\n",
      "b1_c1 (Conv2D)               (None, 60, 71, 8)         224       \n",
      "_________________________________________________________________\n",
      "b1_c2 (Conv2D)               (None, 60, 71, 8)         584       \n",
      "_________________________________________________________________\n",
      "b1_p (MaxPooling2D)          (None, 30, 35, 8)         0         \n",
      "_________________________________________________________________\n",
      "b2_c1 (Conv2D)               (None, 30, 35, 16)        1168      \n",
      "_________________________________________________________________\n",
      "b2_c2 (Conv2D)               (None, 30, 35, 16)        2320      \n",
      "_________________________________________________________________\n",
      "b2_p (MaxPooling2D)          (None, 15, 17, 16)        0         \n",
      "_________________________________________________________________\n",
      "b3_c1 (Conv2D)               (None, 15, 17, 32)        4640      \n",
      "_________________________________________________________________\n",
      "b3_c2 (Conv2D)               (None, 15, 17, 32)        9248      \n",
      "_________________________________________________________________\n",
      "b3_p (MaxPooling2D)          (None, 7, 8, 32)          0         \n",
      "_________________________________________________________________\n",
      "flatten (Flatten)            (None, 1792)              0         \n",
      "_________________________________________________________________\n",
      "FC1 (Dense)                  (None, 128)               229504    \n",
      "_________________________________________________________________\n",
      "FC2 (Dense)                  (None, 64)                8256      \n",
      "_________________________________________________________________\n",
      "Out (Dense)                  (None, 22)                1430      \n",
      "=================================================================\n",
      "Total params: 257,374\n",
      "Trainable params: 257,374\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "Starting to fit model\n",
      "2022-02-01 23:32:44.380105: W tensorflow/core/framework/cpu_allocator_impl.cc:80] Allocation of 1142429760 exceeds 10% of free system memory.\n",
      "Epoch 1/20\n",
      "699/699 [==============================] - 170s 240ms/step - loss: 0.8023 - val_loss: 0.2980\n",
      "Epoch 2/20\n",
      "699/699 [==============================] - 170s 244ms/step - loss: 0.2597 - val_loss: 0.2596\n",
      "Epoch 3/20\n",
      "699/699 [==============================] - 170s 243ms/step - loss: 0.1897 - val_loss: 0.2476\n",
      "Epoch 4/20\n",
      "699/699 [==============================] - 171s 245ms/step - loss: 0.1460 - val_loss: 0.2440\n",
      "Epoch 5/20\n",
      "699/699 [==============================] - 173s 247ms/step - loss: 0.1263 - val_loss: 0.3248\n",
      "Epoch 6/20\n",
      "699/699 [==============================] - 173s 247ms/step - loss: 0.1041 - val_loss: 0.2658\n",
      "Epoch 7/20\n",
      "699/699 [==============================] - 173s 247ms/step - loss: 0.0977 - val_loss: 0.2896\n",
      "Epoch 8/20\n",
      "699/699 [==============================] - 172s 247ms/step - loss: 0.0800 - val_loss: 0.2959\n",
      "Epoch 9/20\n",
      "699/699 [==============================] - 172s 246ms/step - loss: 0.0776 - val_loss: 0.3139\n",
      "Epoch 10/20\n",
      "699/699 [==============================] - 172s 246ms/step - loss: 0.0712 - val_loss: 0.3404\n",
      "Epoch 11/20\n",
      "699/699 [==============================] - 173s 248ms/step - loss: 0.0617 - val_loss: 0.3701\n",
      "Epoch 00011: early stopping\n",
      "fit completed, saving model\n",
      "INFO:tensorflow:Assets written to: doubleConvCNN.CNN/assets\n",
      "Model LeNet5 prepared. Architecture:\n",
      "Model: \"LeNet5\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "In (InputLayer)              [(None, 60, 71, 3)]       0         \n",
      "_________________________________________________________________\n",
      "C1 (Conv2D)                  (None, 56, 67, 6)         456       \n",
      "_________________________________________________________________\n",
      "S2 (AveragePooling2D)        (None, 28, 33, 6)         0         \n",
      "_________________________________________________________________\n",
      "C3 (Conv2D)                  (None, 24, 29, 16)        2416      \n",
      "_________________________________________________________________\n",
      "S4 (AveragePooling2D)        (None, 12, 14, 16)        0         \n",
      "_________________________________________________________________\n",
      "C5 (Conv2D)                  (None, 12, 14, 120)       2040      \n",
      "_________________________________________________________________\n",
      "flatten (Flatten)            (None, 20160)             0         \n",
      "_________________________________________________________________\n",
      "F6 (Dense)                   (None, 84)                1693524   \n",
      "_________________________________________________________________\n",
      "Out (Dense)                  (None, 22)                1870      \n",
      "=================================================================\n",
      "Total params: 1,700,306\n",
      "Trainable params: 1,700,306\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "Starting to fit model\n",
      "2022-02-02 00:04:36.726939: W tensorflow/core/framework/cpu_allocator_impl.cc:80] Allocation of 1142429760 exceeds 10% of free system memory.\n",
      "Epoch 1/20\n",
      "699/699 [==============================] - 109s 155ms/step - loss: 0.6461 - val_loss: 0.3736\n",
      "Epoch 2/20\n",
      "699/699 [==============================] - 106s 152ms/step - loss: 0.3112 - val_loss: 0.3360\n",
      "Epoch 3/20\n",
      "699/699 [==============================] - 106s 152ms/step - loss: 0.2403 - val_loss: 0.3466\n",
      "Epoch 4/20\n",
      "699/699 [==============================] - 106s 152ms/step - loss: 0.1977 - val_loss: 0.2823\n",
      "Epoch 5/20\n",
      "699/699 [==============================] - 89s 127ms/step - loss: 0.1681 - val_loss: 0.2864\n",
      "Epoch 6/20\n",
      "699/699 [==============================] - 63s 90ms/step - loss: 0.1544 - val_loss: 0.3131\n",
      "Epoch 7/20\n",
      "699/699 [==============================] - 62s 89ms/step - loss: 0.1393 - val_loss: 0.2808\n",
      "Epoch 8/20\n",
      "699/699 [==============================] - 62s 89ms/step - loss: 0.1206 - val_loss: 0.2852\n",
      "Epoch 9/20\n",
      "699/699 [==============================] - 62s 88ms/step - loss: 0.1150 - val_loss: 0.3012\n",
      "Epoch 10/20\n",
      "699/699 [==============================] - 62s 89ms/step - loss: 0.1094 - val_loss: 0.2981\n",
      "Epoch 11/20\n",
      "699/699 [==============================] - 62s 88ms/step - loss: 0.1062 - val_loss: 0.3106\n",
      "Epoch 12/20\n",
      "699/699 [==============================] - 62s 89ms/step - loss: 0.1060 - val_loss: 0.3271\n",
      "Epoch 13/20\n",
      "699/699 [==============================] - 61s 88ms/step - loss: 0.0998 - val_loss: 0.3227\n",
      "Epoch 14/20\n",
      "699/699 [==============================] - 62s 88ms/step - loss: 0.0889 - val_loss: 0.3320\n",
      "Epoch 00014: early stopping\n",
      "fit completed, saving model\n",
      "WARNING:absl:Function `_wrapped_model` contains input name(s) In with unsupported characters which will be renamed to in in the SavedModel.\n",
      "INFO:tensorflow:Assets written to: LeNet5.CNN/assets\n",
      "INFO:tensorflow:Assets written to: LeNet5.CNN/assets\n",
      "Model singleConvDroputCNN prepared. Architecture:\n",
      "Model: \"singleConvDroputCNN\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "input_3 (InputLayer)         [(None, 60, 71, 3)]       0         \n",
      "_________________________________________________________________\n",
      "b1_c1 (Conv2D)               (None, 60, 71, 8)         224       \n",
      "_________________________________________________________________\n",
      "b1_p (MaxPooling2D)          (None, 30, 35, 8)         0         \n",
      "_________________________________________________________________\n",
      "b2_c1 (Conv2D)               (None, 30, 35, 16)        1168      \n",
      "_________________________________________________________________\n",
      "b2_p (MaxPooling2D)          (None, 15, 17, 16)        0         \n",
      "_________________________________________________________________\n",
      "b3_c1 (Conv2D)               (None, 15, 17, 32)        4640      \n",
      "_________________________________________________________________\n",
      "b3_p (MaxPooling2D)          (None, 7, 8, 32)          0         \n",
      "_________________________________________________________________\n",
      "flatten (Flatten)            (None, 1792)              0         \n",
      "_________________________________________________________________\n",
      "dropout (Dropout)            (None, 1792)              0         \n",
      "_________________________________________________________________\n",
      "FC1 (Dense)                  (None, 128)               229504    \n",
      "_________________________________________________________________\n",
      "FC2 (Dense)                  (None, 64)                8256      \n",
      "_________________________________________________________________\n",
      "Out (Dense)                  (None, 22)                1430      \n",
      "=================================================================\n",
      "Total params: 245,222\n",
      "Trainable params: 245,222\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "Starting to fit model\n",
      "2022-02-02 00:23:02.592187: W tensorflow/core/framework/cpu_allocator_impl.cc:80] Allocation of 1142429760 exceeds 10% of free system memory.\n",
      "Epoch 1/20\n",
      "699/699 [==============================] - 52s 72ms/step - loss: 0.9002 - val_loss: 0.3688\n",
      "Epoch 2/20\n",
      "699/699 [==============================] - 43s 61ms/step - loss: 0.3482 - val_loss: 0.2544\n",
      "Epoch 3/20\n",
      "699/699 [==============================] - 45s 64ms/step - loss: 0.2649 - val_loss: 0.2436\n",
      "Epoch 4/20\n",
      "699/699 [==============================] - 44s 63ms/step - loss: 0.2303 - val_loss: 0.2302\n",
      "Epoch 5/20\n",
      "699/699 [==============================] - 44s 64ms/step - loss: 0.2037 - val_loss: 0.2308\n",
      "Epoch 6/20\n",
      "699/699 [==============================] - 45s 64ms/step - loss: 0.1800 - val_loss: 0.2282\n",
      "Epoch 7/20\n",
      "699/699 [==============================] - 44s 63ms/step - loss: 0.1756 - val_loss: 0.2235\n",
      "Epoch 8/20\n",
      "699/699 [==============================] - 45s 65ms/step - loss: 0.1573 - val_loss: 0.2310\n",
      "Epoch 9/20\n",
      "699/699 [==============================] - 44s 63ms/step - loss: 0.1471 - val_loss: 0.2309\n",
      "Epoch 10/20\n",
      "699/699 [==============================] - 45s 65ms/step - loss: 0.1389 - val_loss: 0.2330\n",
      "Epoch 11/20\n",
      "699/699 [==============================] - 44s 63ms/step - loss: 0.1321 - val_loss: 0.2333\n",
      "Epoch 12/20\n",
      "699/699 [==============================] - 44s 64ms/step - loss: 0.1241 - val_loss: 0.2401\n",
      "Epoch 13/20\n",
      "699/699 [==============================] - 45s 65ms/step - loss: 0.1215 - val_loss: 0.2343\n",
      "Epoch 14/20\n",
      "699/699 [==============================] - 45s 64ms/step - loss: 0.1159 - val_loss: 0.2317\n",
      "Epoch 00014: early stopping\n",
      "fit completed, saving model\n",
      "INFO:tensorflow:Assets written to: singleConvDroputCNN.CNN/assets\n",
      "INFO:tensorflow:Assets written to: singleConvDroputCNN.CNN/assets\n",
      "Model doubleConvDroputCNN prepared. Architecture:\n",
      "Model: \"doubleConvDroputCNN\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "input_4 (InputLayer)         [(None, 60, 71, 3)]       0         \n",
      "_________________________________________________________________\n",
      "b1_c1 (Conv2D)               (None, 60, 71, 8)         224       \n",
      "_________________________________________________________________\n",
      "b1_c2 (Conv2D)               (None, 60, 71, 8)         584       \n",
      "_________________________________________________________________\n",
      "b1_p (MaxPooling2D)          (None, 30, 35, 8)         0         \n",
      "_________________________________________________________________\n",
      "b2_c1 (Conv2D)               (None, 30, 35, 16)        1168      \n",
      "_________________________________________________________________\n",
      "b2_c2 (Conv2D)               (None, 30, 35, 16)        2320      \n",
      "_________________________________________________________________\n",
      "b2_p (MaxPooling2D)          (None, 15, 17, 16)        0         \n",
      "_________________________________________________________________\n",
      "b3_c1 (Conv2D)               (None, 15, 17, 32)        4640      \n",
      "_________________________________________________________________\n",
      "b3_c2 (Conv2D)               (None, 15, 17, 32)        9248      \n",
      "_________________________________________________________________\n",
      "b3_p (MaxPooling2D)          (None, 7, 8, 32)          0         \n",
      "_________________________________________________________________\n",
      "flatten (Flatten)            (None, 1792)              0         \n",
      "_________________________________________________________________\n",
      "dropout_1 (Dropout)          (None, 1792)              0         \n",
      "_________________________________________________________________\n",
      "FC1 (Dense)                  (None, 128)               229504    \n",
      "_________________________________________________________________\n",
      "FC2 (Dense)                  (None, 64)                8256      \n",
      "_________________________________________________________________\n",
      "Out (Dense)                  (None, 22)                1430      \n",
      "=================================================================\n",
      "Total params: 257,374\n",
      "Trainable params: 257,374\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "Starting to fit model\n",
      "2022-02-02 00:34:16.443565: W tensorflow/core/framework/cpu_allocator_impl.cc:80] Allocation of 1142429760 exceeds 10% of free system memory.\n",
      "Epoch 1/20\n",
      "699/699 [==============================] - 104s 144ms/step - loss: 0.9636 - val_loss: 0.3365\n",
      "Epoch 2/20\n",
      "699/699 [==============================] - 99s 141ms/step - loss: 0.3310 - val_loss: 0.2718\n",
      "Epoch 3/20\n",
      "699/699 [==============================] - 98s 140ms/step - loss: 0.2594 - val_loss: 0.2622\n",
      "Epoch 4/20\n",
      "699/699 [==============================] - 99s 142ms/step - loss: 0.2185 - val_loss: 0.2152\n",
      "Epoch 5/20\n",
      "699/699 [==============================] - 98s 140ms/step - loss: 0.1890 - val_loss: 0.2438\n",
      "Epoch 6/20\n",
      "699/699 [==============================] - 98s 141ms/step - loss: 0.1709 - val_loss: 0.2303\n",
      "Epoch 7/20\n",
      "699/699 [==============================] - 98s 140ms/step - loss: 0.1583 - val_loss: 0.2065\n",
      "Epoch 8/20\n",
      "699/699 [==============================] - 98s 140ms/step - loss: 0.1430 - val_loss: 0.2252\n",
      "Epoch 9/20\n",
      "699/699 [==============================] - 99s 142ms/step - loss: 0.1327 - val_loss: 0.2665\n",
      "Epoch 10/20\n",
      "699/699 [==============================] - 97s 139ms/step - loss: 0.1319 - val_loss: 0.2469\n",
      "Epoch 11/20\n",
      "699/699 [==============================] - 99s 141ms/step - loss: 0.1202 - val_loss: 0.2200\n",
      "Epoch 12/20\n",
      "699/699 [==============================] - 98s 140ms/step - loss: 0.1123 - val_loss: 0.2478\n",
      "Epoch 13/20\n",
      "699/699 [==============================] - 98s 140ms/step - loss: 0.1047 - val_loss: 0.2507\n",
      "Epoch 14/20\n",
      "699/699 [==============================] - 99s 141ms/step - loss: 0.1070 - val_loss: 0.2839\n",
      "Epoch 00014: early stopping\n",
      "fit completed, saving model\n",
      "INFO:tensorflow:Assets written to: doubleConvDroputCNN.CNN/assets\n",
      "INFO:tensorflow:Assets written to: doubleConvDroputCNN.CNN/assets\n",
      "Model LeNet5_Dropout prepared. Architecture:\n",
      "Model: \"LeNet5_Dropout\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "In (InputLayer)              [(None, 60, 71, 3)]       0         \n",
      "_________________________________________________________________\n",
      "C1 (Conv2D)                  (None, 56, 67, 6)         456       \n",
      "_________________________________________________________________\n",
      "S2 (AveragePooling2D)        (None, 28, 33, 6)         0         \n",
      "_________________________________________________________________\n",
      "C3 (Conv2D)                  (None, 24, 29, 16)        2416      \n",
      "_________________________________________________________________\n",
      "S4 (AveragePooling2D)        (None, 12, 14, 16)        0         \n",
      "_________________________________________________________________\n",
      "C5 (Conv2D)                  (None, 12, 14, 120)       2040      \n",
      "_________________________________________________________________\n",
      "flatten_1 (Flatten)          (None, 20160)             0         \n",
      "_________________________________________________________________\n",
      "dropout_2 (Dropout)          (None, 20160)             0         \n",
      "_________________________________________________________________\n",
      "F6 (Dense)                   (None, 84)                1693524   \n",
      "_________________________________________________________________\n",
      "Out (Dense)                  (None, 22)                1870      \n",
      "=================================================================\n",
      "Total params: 1,700,306\n",
      "Trainable params: 1,700,306\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "Starting to fit model\n",
      "Epoch 1/20\n",
      "699/699 [==============================] - 70s 97ms/step - loss: 0.7117 - val_loss: 0.3902\n",
      "Epoch 2/20\n",
      "699/699 [==============================] - 65s 93ms/step - loss: 0.3265 - val_loss: 0.3339\n",
      "Epoch 3/20\n",
      "699/699 [==============================] - 70s 101ms/step - loss: 0.2517 - val_loss: 0.3064\n",
      "Epoch 4/20\n",
      "699/699 [==============================] - 67s 95ms/step - loss: 0.2131 - val_loss: 0.2813\n",
      "Epoch 5/20\n",
      "699/699 [==============================] - 66s 95ms/step - loss: 0.1818 - val_loss: 0.2887\n",
      "Epoch 6/20\n",
      "699/699 [==============================] - 68s 98ms/step - loss: 0.1645 - val_loss: 0.2742\n",
      "Epoch 7/20\n",
      "699/699 [==============================] - 67s 95ms/step - loss: 0.1545 - val_loss: 0.2913\n",
      "Epoch 8/20\n",
      "699/699 [==============================] - 67s 95ms/step - loss: 0.1422 - val_loss: 0.2860\n",
      "Epoch 9/20\n",
      "699/699 [==============================] - 66s 94ms/step - loss: 0.1382 - val_loss: 0.2819\n",
      "Epoch 10/20\n",
      "699/699 [==============================] - 68s 98ms/step - loss: 0.1264 - val_loss: 0.3016\n",
      "Epoch 11/20\n",
      "699/699 [==============================] - 66s 95ms/step - loss: 0.1212 - val_loss: 0.3032\n",
      "Epoch 12/20\n",
      "699/699 [==============================] - 66s 95ms/step - loss: 0.1261 - val_loss: 0.3147\n",
      "Epoch 13/20\n",
      "699/699 [==============================] - 66s 95ms/step - loss: 0.1084 - val_loss: 0.2862\n",
      "Epoch 00013: early stopping\n",
      "fit completed, saving model\n",
      "WARNING:absl:Function `_wrapped_model` contains input name(s) In with unsupported characters which will be renamed to in in the SavedModel.\n",
      "INFO:tensorflow:Assets written to: LeNet5_Dropout.CNN/assets\n",
      "INFO:tensorflow:Assets written to: LeNet5_Dropout.CNN/assets\n"
     ]
    }
   ],
   "source": [
    "#training\n",
    "from tensorflow import keras\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Dense, InputLayer\n",
    "import os\n",
    "import pickle\n",
    "\n",
    "for model in models_totry:\n",
    "\n",
    "        #compiling model with Adam optimizer\n",
    "        model.compile(keras.optimizers.Adam(), loss='categorical_crossentropy')\n",
    "\n",
    "        print(f\"Model {model.name} prepared. Architecture:\")\n",
    "        model.summary()\n",
    "\n",
    "        #preparing callback for early stopping\n",
    "        es = keras.callbacks.EarlyStopping(monitor='val_loss', mode='min', verbose=1, patience=7)\n",
    "\n",
    "        #start training\n",
    "        print(\"Starting to fit model\")\n",
    "        history = model.fit(x=x_train, y=y_train_mrx, epochs=20, verbose=True, shuffle=True, validation_data=(x_test, y_test_mrx), callbacks=[es])\n",
    "    \n",
    "        #saving trained model\n",
    "        print(\"fit completed, saving model\")\n",
    "\n",
    "        #preparaing saving dir\n",
    "        saving_dir = f\"{model.name}.CNN\"\n",
    "        os.mkdir(saving_dir)\n",
    "        \n",
    "        #saving history\n",
    "        with open(f\"{saving_dir}/history.hist\", 'wb') as file_h:\n",
    "            pickle.dump(history.history, file_h)\n",
    "            \n",
    "        #saving model\n",
    "        model.save(saving_dir)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Results\n",
    "\n",
    "In this section I will compare the trained models.\n",
    "\n",
    "First of all I will show how the loss function computed on the training and on the test set changed during the training process. This images will provide a first impression of the models' performances and are useful to qualitatively asses if the networks were trained correctly.\n",
    "\n",
    "I will then compute and show the normalized confusion matrices of all the trained models.\n",
    "\n",
    "Then I will compute the 4 metrics computed also for all the other methods: Accuracy, Balanced accuracy, F1-score and MCC.\n",
    "\n",
    "Finally, based on this information I will select the better model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x360 with 6 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "#plot cross entropy minimization during training\n",
    "import pickle\n",
    "import matplotlib.pyplot as plt\n",
    "import re\n",
    "import os\n",
    "\n",
    "mod_name_re = re.compile(\".*\\.CNN$\")\n",
    "models = [s for s in os.listdir('./') if mod_name_re.match(s)]\n",
    "fig, axs = plt.subplots(2,int(len(models)/2), figsize=(10, 5), constrained_layout=True)\n",
    "\n",
    "axs = axs.flatten()\n",
    "\n",
    "for j, model in enumerate(models):\n",
    "    with open(f\"./{model}/history.hist\", 'rb') as hist:\n",
    "        history = pickle.load(hist)\n",
    "    axs[j].plot(history['loss'], label='training set')\n",
    "    axs[j].plot(history['val_loss'], label='test set')\n",
    "    axs[j].set_xscale('log')\n",
    "    axs[j].set_xlabel('Epochs')\n",
    "    axs[j].set_ylabel('CCE')\n",
    "    axs[j].set_title(f\"{model}\")\n",
    "    axs[j].legend()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The trainings of all of the models were early stopped after about 10 epochs. From the loss history it is possible to observe that, while its value computed on the training set gets gradually lower, the value computed on the test set witness little or no reduction with respect to the value reached after the first epoch. \n",
    "These models easily overfit the data."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "opening model doubleConvDroputCNN.CNN\n",
      "opening model singleConvDroputCNN.CNN\n",
      "opening model singleConvCNN.CNN\n",
      "opening model doubleConvCNN.CNN\n",
      "opening model LeNet5_Dropout.CNN\n",
      "opening model LeNet5.CNN\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  0%|                                                       | 0/6 [00:00<?, ?it/s]2022-02-03 15:51:14.572403: W tensorflow/core/framework/cpu_allocator_impl.cc:80] Allocation of 241286400 exceeds 10% of free system memory.\n",
      "2022-02-03 15:51:15.750803: I tensorflow/compiler/mlir/mlir_graph_optimization_pass.cc:176] None of the MLIR Optimization Passes are enabled (registered 2)\n",
      "2022-02-03 15:51:15.800155: I tensorflow/core/platform/profile_utils/cpu_utils.cc:114] CPU Frequency: 2594540000 Hz\n",
      " 33%|███████████████▋                               | 2/6 [00:07<00:14,  3.50s/it]2022-02-03 15:51:21.297490: W tensorflow/core/framework/cpu_allocator_impl.cc:80] Allocation of 241286400 exceeds 10% of free system memory.\n",
      " 50%|███████████████████████▌                       | 3/6 [00:10<00:10,  3.62s/it]2022-02-03 15:51:24.620132: W tensorflow/core/framework/cpu_allocator_impl.cc:80] Allocation of 241286400 exceeds 10% of free system memory.\n",
      " 67%|███████████████████████████████▎               | 4/6 [00:14<00:07,  3.50s/it]2022-02-03 15:51:27.893568: W tensorflow/core/framework/cpu_allocator_impl.cc:80] Allocation of 241286400 exceeds 10% of free system memory.\n",
      " 83%|███████████████████████████████████████▏       | 5/6 [00:19<00:03,  4.00s/it]2022-02-03 15:51:32.899865: W tensorflow/core/framework/cpu_allocator_impl.cc:80] Allocation of 241286400 exceeds 10% of free system memory.\n",
      "100%|███████████████████████████████████████████████| 6/6 [00:28<00:00,  4.68s/it]\n",
      "100%|█████████████████████████████████████████████| 6/6 [00:00<00:00, 1724.75it/s]\n"
     ]
    }
   ],
   "source": [
    "#loading all the models in a dataframe\n",
    "import pandas as pd\n",
    "import tensorflow.keras as keras\n",
    "results_df = pd.DataFrame()\n",
    "\n",
    "for i, model in enumerate(models):\n",
    "        print(f'opening model {model}')\n",
    "        kmod = keras.models.load_model(f'./{model}')\n",
    "        new_entry = {'model_name': model, 'model': kmod}\n",
    "        results_df = results_df.append(new_entry, ignore_index=True)\n",
    "\n",
    "#importing tqdm for progress bar\n",
    "from tqdm import tqdm\n",
    "tqdm.pandas()\n",
    "\n",
    "#computing the predictions of each model on the test set\n",
    "results_df['predicted_y_p'] = results_df['model'].progress_apply(lambda classif: classif.predict(x_test))\n",
    "results_df['predicted_y'] = results_df['predicted_y_p'].progress_apply(lambda p: p.argmax(axis=1))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████████████████████████████████████████| 6/6 [00:00<00:00, 170.91it/s]\n",
      "100%|██████████████████████████████████████████████| 6/6 [00:00<00:00, 178.81it/s]\n"
     ]
    }
   ],
   "source": [
    "from sklearn.metrics import confusion_matrix\n",
    "\n",
    "#computing confusion matrix for each model\n",
    "results_df['cmatrix'] = results_df['predicted_y'].progress_apply(lambda y_pred: confusion_matrix(y_test, y_pred))\n",
    "\n",
    "#computing, for each model, the confusion matrix \n",
    "#with each element normalized with the number of elements \n",
    "#belonging to the same true class\n",
    "results_df['cmatrix_truen'] = results_df['predicted_y'].progress_apply(lambda y_pred: confusion_matrix(y_test, y_pred, normalize='true'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['doubleConvDroputCNN.CNN',\n",
       " 'singleConvDroputCNN.CNN',\n",
       " 'singleConvCNN.CNN',\n",
       " 'doubleConvCNN.CNN',\n",
       " 'LeNet5_Dropout.CNN',\n",
       " 'LeNet5.CNN']"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "models"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/tmp/ipykernel_455950/4240995028.py:30: UserWarning: Matplotlib is currently using module://matplotlib_inline.backend_inline, which is a non-GUI backend, so cannot show the figure.\n",
      "  fig.show()\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1008x720 with 7 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "#plotting confusion matrices\n",
    "fig, axs = plt.subplots(2, int(len(models)/2), figsize=(14, 10), constrained_layout=True)\n",
    "\n",
    "for i, ax in enumerate(axs.flatten()):\n",
    "        cmat = results_df['cmatrix_truen'].iloc[i]\n",
    "        im = ax.imshow(cmat, cmap='Blues')\n",
    "        ax.set_title(f\"{results_df.iloc[i]['model_name']}\")\n",
    "        ax.set_xticks(range(0,22))\n",
    "        ax.set_yticks(range(0,22))\n",
    "        ax.tick_params(axis='both', which='major', labelsize=7)\n",
    "\n",
    "fig.suptitle('Confusion matrices normalized with true number of elements in each class', fontsize=20)\n",
    "fig.supxlabel('predicted class')\n",
    "fig.supylabel('actual class')\n",
    "\n",
    "fig.colorbar(im,  ax=axs.ravel().tolist())\n",
    "\n",
    "#generate class legend\n",
    "fig.text(1.05, 0.92, \"Class labels and frequency\\n in the test set\", fontsize=15)\n",
    "class_labels = u.get_class_labels()\n",
    "class_labels['test_perc'] = class_labels['n_test']*100/class_labels['n_test'].sum()\n",
    "classes=\"\"\n",
    "perc=\"\"\n",
    "for i, row in class_labels.iterrows():\n",
    "    classes = classes + f\"{row['class_id']:>2}:  {row['class'].replace('_', ' '):<25}\\n\"\n",
    "    perc = perc + f\"{row['test_perc']:>5.2f}%\\n\"\n",
    "\n",
    "fig.text(1.05, 0.50, classes, fontsize=11)\n",
    "fig.text(1.3, 0.50, perc, fontsize=11) \n",
    "fig.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The confusion matrices in the figure above give a first impression of the models' performances. \n",
    "They seem similar to the ones computed for the previously used methods. The misclassifcation problem of 'Paired Doves' elements remains.\n",
    "Furthermore, the LeNet5 architecture seems to provide worse results than the other models even with the addition of the dropout layer. This might be due to the use of the saturating `tanh` activation function."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|█████████████████████████████████████████████| 6/6 [00:00<00:00, 9098.27it/s]\n",
      "100%|█████████████████████████████████████████████| 6/6 [00:00<00:00, 4712.70it/s]\n",
      "100%|█████████████████████████████████████████████| 6/6 [00:00<00:00, 1703.85it/s]\n",
      "100%|█████████████████████████████████████████████| 6/6 [00:00<00:00, 3313.04it/s]\n"
     ]
    }
   ],
   "source": [
    "#computing metrics\n",
    "\n",
    "#Accuracy and  balanced accuracy\n",
    "def compute_balacc(cmat):\n",
    "    bacc = 0\n",
    "    for i, row in enumerate(cmat):\n",
    "        bacc += row[i]/row.sum()\n",
    "    return bacc/len(cmat)\n",
    "results_df['accuracy'] = results_df['cmatrix'].progress_apply(lambda cmat: cmat.trace()/cmat.sum())\n",
    "results_df['balanced_accuracy'] = results_df['cmatrix'].progress_apply(lambda cmat: compute_balacc(cmat))\n",
    "\n",
    "#F1-score and MCC\n",
    "#F1 score\n",
    "def compute_macroF1(cmat):\n",
    "    mprec = 0\n",
    "    mrec = 0\n",
    "    #compute macro precision and macro recall\n",
    "    for i in range(0, len(cmat)):\n",
    "        mprec += (cmat[i][i]/(cmat[:,i].sum()+1e-20))\n",
    "        mrec  += (cmat[i][i]/(cmat[i].sum()+1e-20))\n",
    "   \n",
    "    mprec /= len(cmat)\n",
    "    mrec  /= len(cmat)\n",
    "\n",
    "    return 2*(mprec*mrec)/(mprec+mrec)\n",
    "\n",
    "def compute_MCC(cmat):\n",
    "    c = cmat.trace()\n",
    "    s = cmat.sum()\n",
    "    p = [cmat[:,k].sum()+1e-20 for k in range(0, len(cmat))]\n",
    "    t = [cmat[k].sum() for k in range(0, len(cmat))]\n",
    "\n",
    "    mcc = (c*s - np.dot(p,t))/np.sqrt((s**2-np.dot(p,p))*(s**2-np.dot(t,t)))\n",
    "    return mcc\n",
    "\n",
    "results_df['macroF1'] = results_df['cmatrix'].progress_apply(lambda cmat: compute_macroF1(cmat))\n",
    "results_df['MCC'] = results_df['cmatrix'].progress_apply(lambda cmat: compute_MCC(cmat))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<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>model_name</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>balanced_accuracy</th>\n",
       "      <th>macroF1</th>\n",
       "      <th>MCC</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>doubleConvDroputCNN.CNN</td>\n",
       "      <td>0.9468</td>\n",
       "      <td>0.8992</td>\n",
       "      <td>0.9026</td>\n",
       "      <td>0.9414</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>singleConvDroputCNN.CNN</td>\n",
       "      <td>0.9564</td>\n",
       "      <td>0.9021</td>\n",
       "      <td>0.9174</td>\n",
       "      <td>0.9518</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>singleConvCNN.CNN</td>\n",
       "      <td>0.9462</td>\n",
       "      <td>0.8892</td>\n",
       "      <td>0.8991</td>\n",
       "      <td>0.9405</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>doubleConvCNN.CNN</td>\n",
       "      <td>0.9377</td>\n",
       "      <td>0.8664</td>\n",
       "      <td>0.8817</td>\n",
       "      <td>0.9312</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>LeNet5_Dropout.CNN</td>\n",
       "      <td>0.9271</td>\n",
       "      <td>0.8406</td>\n",
       "      <td>0.8666</td>\n",
       "      <td>0.9194</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>LeNet5.CNN</td>\n",
       "      <td>0.9227</td>\n",
       "      <td>0.8258</td>\n",
       "      <td>0.8545</td>\n",
       "      <td>0.9145</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                model_name  accuracy  balanced_accuracy  macroF1     MCC\n",
       "0  doubleConvDroputCNN.CNN    0.9468             0.8992   0.9026  0.9414\n",
       "1  singleConvDroputCNN.CNN    0.9564             0.9021   0.9174  0.9518\n",
       "2        singleConvCNN.CNN    0.9462             0.8892   0.8991  0.9405\n",
       "3        doubleConvCNN.CNN    0.9377             0.8664   0.8817  0.9312\n",
       "4       LeNet5_Dropout.CNN    0.9271             0.8406   0.8666  0.9194\n",
       "5               LeNet5.CNN    0.9227             0.8258   0.8545  0.9145"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "results_df[['model_name', 'accuracy', 'balanced_accuracy', 'macroF1', 'MCC']].round(4)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "This table presents the metrics of each model. \n",
    "\n",
    "All the architectures are observed to benefit from the introduction of the dropout layer with an increment of 1 to 2 percentage points for all of the computed metrics.\n",
    "\n",
    "As already mentioned the LeNet5 architecture provided the worse results. The singleConv model, with less trainable parameters of all the others, has the greatest scores."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Conclusion\n",
    "\n",
    "The convolutional neural network that provided the best results was the `singleConvDropoutCNN`.\n",
    "\n",
    "From its application to the test set it obtained the following scores.\n",
    "\n",
    "    Accuracy: 95.64%\n",
    "    Balanced Accuracy: 90.21%\n",
    "    F1-score: 91.74%\n",
    "    MCC: 95.18%"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## References\n",
    "\n",
    "_Conv2D layer - keras documentation_, Available at\n",
    "https://keras.io/api/layers/convolution_layers/convolution2d/\n",
    "\n",
    "Géron, Aurélien. _Hands-on machine learning with Scikit-Learn, Keras, and TensorFlow: Concepts, tools, and techniques to build intelligent systems._ O'Reilly Media, 2019.\n",
    "\n",
    "Google developers. _Machine learning glossary_, Available at: https://developers.google.com/machine-learning/glossary/#convolutional_neural_network\n",
    "\n",
    "LeCun, Yann, et al. _Backpropagation applied to handwritten zip code recognition._ Neural computation 1.4 (1989): 541-551."
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3.9.5 64-bit",
   "language": "python",
   "name": "python39564bit80cf730d14af42a98fb2c59e9ca33086"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.9.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}