examples/models-usages/cnn-classification/cnn_classification_mnist.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Simple MNIST multiclass classification (using CNN)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "tags": []
   },
   "source": [
    "## Imports"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2024-11-17T11:11:52.331449300Z",
     "start_time": "2024-11-17T11:11:52.281453600Z"
    }
   },
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "from tensorflow.keras.datasets import mnist\n",
    "\n",
    "from neuralnetlib.activations import ReLU\n",
    "from neuralnetlib.layers import Input, Conv2D, MaxPooling2D, Flatten, Dense, Activation\n",
    "from neuralnetlib.models import Sequential\n",
    "from neuralnetlib.preprocessing import one_hot_encode\n",
    "from neuralnetlib.metrics import accuracy_score, recall_score, f1_score \n",
    "from neuralnetlib.utils import train_test_split"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 1. Loading a dataset (in this case, MNIST)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2024-11-17T11:11:52.444357500Z",
     "start_time": "2024-11-17T11:11:52.289970400Z"
    }
   },
   "outputs": [],
   "source": [
    "(x_train, y_train), (x_test, y_test) = mnist.load_data()\n",
    "x_test, x_val, y_test, y_val = train_test_split(x_test, y_test, test_size=0.2, random_state=42)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 2. Preprocessing"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2024-11-17T11:11:52.554589700Z",
     "start_time": "2024-11-17T11:11:52.445357900Z"
    }
   },
   "outputs": [],
   "source": [
    "x_train = x_train.reshape(-1, 28, 28, 1) / 255.0  # Normalization and reshaping of the images for CNN\n",
    "x_test = x_test.reshape(-1, 28, 28, 1) / 255.0    # Normalization and reshaping of the images for CNN\n",
    "x_val = x_val.reshape(-1, 28, 28, 1) / 255.0      # Normalization and reshaping of the images for CNN\n",
    "\n",
    "y_train = one_hot_encode(y_train, num_classes=10)  # One-hot encoding of the labels\n",
    "y_test = one_hot_encode(y_test, num_classes=10)  # One-hot encoding of the labels\n",
    "y_val = one_hot_encode(y_val, num_classes=10)  # One-hot encoding of the labels"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 3. Model definition"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2024-11-17T11:11:52.570116100Z",
     "start_time": "2024-11-17T11:11:52.557094400Z"
    }
   },
   "outputs": [],
   "source": [
    "model = Sequential()\n",
    "model.add(Input(input_shape=(28, 28, 1)))\n",
    "model.add(Conv2D(filters=8, kernel_size=(3, 3), random_state=42))\n",
    "model.add(Activation(ReLU()))\n",
    "model.add(MaxPooling2D(pool_size=2))\n",
    "model.add(Conv2D(filters=16, kernel_size=(3, 3), random_state=42))\n",
    "model.add(Activation(ReLU()))\n",
    "model.add(MaxPooling2D(pool_size=(2, 2)))\n",
    "model.add(Flatten())\n",
    "model.add(Dense(32, random_state=42))\n",
    "model.add(Activation(ReLU()))\n",
    "model.add(Dense(10, random_state=42, activation=\"softmax\"))  # Yeah, you can also use strings for the activation functions, or directly the class"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 4. Model compilation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2024-11-17T11:11:52.614690600Z",
     "start_time": "2024-11-17T11:11:52.570116100Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Sequential(gradient_clip_threshold=5.0, enable_padding=False, padding_size=32, random_state=1733508031486758100)\n",
      "-------------------------------------------------\n",
      "Layer 1: Input(input_shape=(28, 28, 1))\n",
      "Layer 2: Conv2D(num_filters=8, kernel_size=(3, 3), strides=(1, 1), padding=valid)\n",
      "Layer 3: Activation(ReLU)\n",
      "Layer 4: MaxPooling2D(pool_size=(2, 2), strides=(2, 2), padding=valid)\n",
      "Layer 5: Conv2D(num_filters=16, kernel_size=(3, 3), strides=(1, 1), padding=valid)\n",
      "Layer 6: Activation(ReLU)\n",
      "Layer 7: MaxPooling2D(pool_size=(2, 2), strides=(2, 2), padding=valid)\n",
      "Layer 8: Flatten\n",
      "Layer 9: Dense(units=32)\n",
      "Layer 10: Activation(ReLU)\n",
      "Layer 11: Dense(units=10)\n",
      "Layer 12: Activation(Softmax)\n",
      "-------------------------------------------------\n",
      "Loss function: CategoricalCrossentropy\n",
      "Optimizer: Adam(learning_rate=0.001, beta_1=0.9, beta_2=0.999, epsilon=1e-08, clip_norm=None, clip_value=None)\n",
      "-------------------------------------------------\n",
      "\n"
     ]
    }
   ],
   "source": [
    "model.compile(loss_function=\"cce\", optimizer=\"adam\")  # You can also use strings for the loss function and the optimizer\n",
    "\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 5. Model training"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "ExecuteTime": {
     "start_time": "2024-11-17T11:11:52.585642Z"
    },
    "is_executing": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[==============================] 100% Epoch 1/10 - 35.50s - loss: 0.3256 - accuracy: 0.8995 - val_loss: 0.1242 - val_accuracy: 0.9630\n",
      "[==============================] 100% Epoch 2/10 - 33.76s - loss: 0.1166 - accuracy: 0.9651 - val_loss: 0.0835 - val_accuracy: 0.9739\n",
      "[==============================] 100% Epoch 3/10 - 32.98s - loss: 0.0918 - accuracy: 0.9727 - val_loss: 0.0693 - val_accuracy: 0.9782\n",
      "[==============================] 100% Epoch 4/10 - 34.40s - loss: 0.0792 - accuracy: 0.9768 - val_loss: 0.0653 - val_accuracy: 0.9804\n",
      "[==============================] 100% Epoch 5/10 - 34.27s - loss: 0.0711 - accuracy: 0.9798 - val_loss: 0.0666 - val_accuracy: 0.9812\n",
      "[==============================] 100% Epoch 6/10 - 35.14s - loss: 0.0646 - accuracy: 0.9821 - val_loss: 0.0653 - val_accuracy: 0.9820\n",
      "[==============================] 100% Epoch 7/10 - 36.32s - loss: 0.0600 - accuracy: 0.9845 - val_loss: 0.0693 - val_accuracy: 0.9816\n",
      "[==============================] 100% Epoch 8/10 - 34.42s - loss: 0.0563 - accuracy: 0.9861 - val_loss: 0.0740 - val_accuracy: 0.9815\n",
      "[==============================] 100% Epoch 9/10 - 34.39s - loss: 0.0536 - accuracy: 0.9871 - val_loss: 0.0734 - val_accuracy: 0.9806\n",
      "[==============================] 100% Epoch 10/10 - 34.89s - loss: 0.0518 - accuracy: 0.9874 - val_loss: 0.0738 - val_accuracy: 0.9831\n",
      "\n"
     ]
    },
    {
     "data": {
      "text/plain": []
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.fit(x_train, y_train, epochs=10, batch_size=64, metrics=[\n",
    "            \"accuracy\"], random_state=42, validation_data=(x_test, y_test))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 6. Model evaluation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "is_executing": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test loss: 0.09496325636635701\n"
     ]
    }
   ],
   "source": [
    "loss, _ = model.evaluate(x_val, y_val)\n",
    "print(f'Test loss: {loss}')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 7. Model prediction"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "is_executing": true
   },
   "outputs": [],
   "source": [
    "y_pred = model.predict(x_val)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 8. Printing some metrics"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "is_executing": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "accuracy: 0.9765\n",
      "f1_score: 0.9761285749047082\n",
      "recall_score 0.9758597470665299\n"
     ]
    }
   ],
   "source": [
    "print(\"accuracy:\", accuracy_score(y_pred, y_val))\n",
    "print(\"f1_score:\", f1_score(y_pred, y_val))\n",
    "print(\"recall_score\", recall_score(y_pred, y_val))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 9. Plot the first 10 test images, their predicted labels, and the true labels."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "is_executing": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 1000x1000 with 10 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig = plt.figure(figsize=(10, 10))\n",
    "for i in range(10):\n",
    "    ax = fig.add_subplot(5, 2, i + 1, xticks=[], yticks=[])\n",
    "    ax.imshow(x_val[i].reshape(28, 28), cmap='gray')\n",
    "    ax.set_title(f\"Predicted: {np.argmax(y_pred[i])}, Actual: {np.argmax(y_val[i])}\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 10. Save the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "is_executing": true
   },
   "outputs": [],
   "source": [
    "model.save(\"my_mnist_model.npz\")"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.10.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}