{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "13fa2f64",
   "metadata": {},
   "outputs": [],
   "source": [
    "# In this example, we used the TensorFlow library to load the MNIST data,\n",
    "# define an MLP model with three dense layers, compile the model, train it\n",
    "# for 10 epochs, evaluate it on the test set, and make predictions on\n",
    "# the test set. Finally, we plot some examples of the predictions made\n",
    "# by the model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "1c4405f8",
   "metadata": {},
   "outputs": [],
   "source": [
    "import tensorflow as tf\n",
    "from tensorflow import keras\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "(x_train, y_train), (x_test, y_test) = tf.keras.datasets.mnist.load_data()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "1e7e58fb",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Normalize the pixel values to be between 0 and 1\n",
    "x_train = x_train / 255\n",
    "x_test = x_test / 255"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "3d8a7370",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Flatten the 2D images into 1D arrays\n",
    "x_train = x_train.reshape(x_train.shape[0], -1)\n",
    "x_test = x_test.reshape(x_test.shape[0], -1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "c2df0f54",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Convert the labels into one-hot encoded arrays\n",
    "y_train = tf.keras.utils.to_categorical(y_train, num_classes=10)\n",
    "y_test = tf.keras.utils.to_categorical(y_test, num_classes=10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "e2b249ea",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Define the model\n",
    "# The number of parameters depends on the shapes and sizes of the layers.\n",
    "# In the given model, the first layer Dense(512, activation='relu',\n",
    "# input_shape=(784,)) has 784 input nodes and 512 output nodes. Therefore,\n",
    "# the number of parameters in this layer would be (784 * 512) + 512 = 401920,\n",
    "# where the +512 term is for the bias terms.\n",
    "# The second layer also has 512 input nodes and 512 output nodes, which makes\n",
    "# 512512 = 262,144 parameters. The third and last layer has 512 input nodes\n",
    "# and 10 output nodes, which makes 512*10 = 5,120 parameters.\n",
    "model = tf.keras.models.Sequential()\n",
    "model.add(tf.keras.layers.Dense(512, activation='relu', input_shape=(784,)))\n",
    "model.add(tf.keras.layers.Dense(512, activation='relu'))\n",
    "model.add(tf.keras.layers.Dense(10, activation='softmax'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "bab8730a",
   "metadata": {},
   "outputs": [],
   "source": [
    "# over come overfitting by regularization\n",
    "#model = tf.keras.models.Sequential([\n",
    "#    tf.keras.layers.Dense(64, activation='relu', kernel_regularizer=tf.keras.regularizers.l2(0.001),input_shape=(784,)),\n",
    "#    tf.keras.layers.Dense(64, activation='relu', kernel_regularizer=tf.keras.regularizers.l2(0.001)),\n",
    "#    tf.keras.layers.Dense(10, activation='softmax')\n",
    "#])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "e3223c61",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Compile the model\n",
    "model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "51e7758f",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Train the model and record the history\n",
    "history = model.fit(x_train, y_train, epochs=10, batch_size=64, validation_data=(x_test, y_test))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "e3240069",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Get the weights of the Dense layer\n",
    "# plot the weights as a heatmap or image, where the weights are represented\n",
    "# as pixel values.\n",
    "# model.layers[2].get_weights()[0] returns only the weights of the third\n",
    "# layer. If you wanted to get the biases, you would use\n",
    "# model.layers[2].get_weights()[1].\n",
    "dense_weights = model.layers[2].get_weights()[0]\n",
    "\n",
    "# Plot the weights as a heatmap\n",
    "fig, ax = plt.subplots(figsize=(12, 36))\n",
    "im = ax.imshow(dense_weights, cmap='coolwarm')\n",
    "plt.colorbar(im, ax=ax)\n",
    "ax.set_title('Weights in the Output Layer')\n",
    "ax.set_xlabel('Neurons in the Output Layer')\n",
    "ax.set_ylabel('Neurons in the Previous Layer')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "13118686",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Evaluate the model on the test set\n",
    "test_loss, test_acc = model.evaluate(x_test, y_test)\n",
    "print('Test accuracy:', test_acc)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "40f35fe5",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Plot loss and accuracy\n",
    "plt.figure(figsize=(12, 4))\n",
    "\n",
    "# Plot the loss and accuracy for training and validation data\n",
    "plt.subplot(1, 2, 1)\n",
    "plt.plot(history.history['loss'], label='training loss')\n",
    "plt.plot(history.history['val_loss'], label='validation loss')\n",
    "plt.xlabel('Epoch')\n",
    "plt.ylabel('Loss')\n",
    "plt.legend()\n",
    "\n",
    "plt.subplot(1, 2, 2)\n",
    "plt.plot(history.history['accuracy'])\n",
    "plt.plot(history.history['val_accuracy'])\n",
    "plt.xlabel('Epoch')\n",
    "plt.ylabel('Accuracy')\n",
    "plt.legend()\n",
    "\n",
    "plt.show()\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "e882afea",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Plot a confusion matrix of the test set predictions\n",
    "test_preds = np.argmax(model.predict(x_test), axis=1)\n",
    "conf_mat = tf.math.confusion_matrix(y_test.argmax(axis=1), test_preds)\n",
    "plt.imshow(conf_mat, cmap=\"Blues\")\n",
    "plt.xlabel(\"Predicted labels\")\n",
    "plt.ylabel(\"True labels\")\n",
    "plt.xticks(np.arange(10))\n",
    "plt.yticks(np.arange(10))\n",
    "plt.colorbar()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "6176ce1e",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Make predictions on the test set\n",
    "y_pred = model.predict(x_test)\n",
    "y_pred = np.argmax(y_pred, axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "3635ded5",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Plot some examples from the test set and their predictions\n",
    "fig, axes = plt.subplots(4, 4, figsize=(14, 14))\n",
    "for i, ax in enumerate(axes.ravel()):\n",
    "    ax.matshow(x_test[i].reshape(28, 28), cmap='gray')\n",
    "    ax.set_title(\"True: %d\\nPredict: %d\" % (np.argmax(y_test[i]), y_pred[i]))\n",
    "    ax.axis(\"off\")\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "71f1cb93",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.16"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}