build_linear_regression.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Linear Regression"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Install the required libraries\n",
    "```bash\n",
    "pip install numpy pandas matplotlib seaborn scikit-learn tensorflow\n",
    "```"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# !pip install numpy pandas matplotlib seaborn scikit-learn tensorflow"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Load and split the data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import tensorflow as tf\n",
    "import numpy as np\n",
    "from tensorflow.python.framework.convert_to_constants import convert_variables_to_constants_v2\n",
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "# Constants for the linear equation\n",
    "a = 2  # Slope\n",
    "c = 3  # Intercept\n",
    "\n",
    "# Generate dataset with some noise and the model function is -z = ax + c\n",
    "np.random.seed(42)  # For reproducibility\n",
    "x_data = np.linspace(-10, 10, 100)  # Generate 100 data points\n",
    "z_data = a * x_data + c + np.random.normal(0, 1, 100)  # Add some noise\n",
    "\n",
    "# Split dataset into training and validation\n",
    "x_train, x_val, z_train, z_val = train_test_split(x_data, z_data, test_size=0.2, random_state=42)\n",
    "x_test, x_val, z_test, z_val = train_test_split(x_val, z_val, test_size=0.5, random_state=42)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Build the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      " Layer (type)                Output Shape              Param #   \n",
      "=================================================================\n",
      " dense (Dense)               (None, 1)                 2         \n",
      "                                                                 \n",
      "=================================================================\n",
      "Total params: 2 (8.00 Byte)\n",
      "Trainable params: 2 (8.00 Byte)\n",
      "Non-trainable params: 0 (0.00 Byte)\n",
      "_________________________________________________________________\n",
      "Epoch 1/100\n",
      "3/3 [==============================] - 0s 31ms/step - loss: 148.9045 - val_loss: 6.8018\n",
      "Epoch 2/100\n",
      "3/3 [==============================] - 0s 5ms/step - loss: 7.8797 - val_loss: 6.0805\n",
      "Epoch 3/100\n",
      "3/3 [==============================] - 0s 5ms/step - loss: 7.0335 - val_loss: 5.4534\n",
      "Epoch 4/100\n",
      "3/3 [==============================] - 0s 5ms/step - loss: 6.3608 - val_loss: 5.0536\n",
      "Epoch 5/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 5.8199 - val_loss: 4.1015\n",
      "Epoch 6/100\n",
      "3/3 [==============================] - 0s 5ms/step - loss: 5.2837 - val_loss: 3.9972\n",
      "Epoch 7/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 4.6115 - val_loss: 3.3158\n",
      "Epoch 8/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 4.2558 - val_loss: 3.0855\n",
      "Epoch 9/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 3.8448 - val_loss: 2.7301\n",
      "Epoch 10/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 3.5064 - val_loss: 2.4430\n",
      "Epoch 11/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 3.2902 - val_loss: 2.0371\n",
      "Epoch 12/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 2.9784 - val_loss: 1.9553\n",
      "Epoch 13/100\n",
      "3/3 [==============================] - 0s 5ms/step - loss: 2.7046 - val_loss: 2.0533\n",
      "Epoch 14/100\n",
      "3/3 [==============================] - 0s 5ms/step - loss: 2.5448 - val_loss: 1.8106\n",
      "Epoch 15/100\n",
      "3/3 [==============================] - 0s 5ms/step - loss: 2.3215 - val_loss: 1.3898\n",
      "Epoch 16/100\n",
      "3/3 [==============================] - 0s 5ms/step - loss: 2.2233 - val_loss: 1.1982\n",
      "Epoch 17/100\n",
      "3/3 [==============================] - 0s 5ms/step - loss: 2.0777 - val_loss: 1.2237\n",
      "Epoch 18/100\n",
      "3/3 [==============================] - 0s 5ms/step - loss: 1.8910 - val_loss: 1.0504\n",
      "Epoch 19/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 1.8221 - val_loss: 1.0322\n",
      "Epoch 20/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 1.6888 - val_loss: 0.7855\n",
      "Epoch 21/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 1.5961 - val_loss: 0.8982\n",
      "Epoch 22/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 1.5477 - val_loss: 0.9218\n",
      "Epoch 23/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 1.4607 - val_loss: 0.6567\n",
      "Epoch 24/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 1.3658 - val_loss: 0.6501\n",
      "Epoch 25/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 1.3088 - val_loss: 0.5586\n",
      "Epoch 26/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 1.2725 - val_loss: 0.5022\n",
      "Epoch 27/100\n",
      "3/3 [==============================] - 0s 5ms/step - loss: 1.2115 - val_loss: 0.5164\n",
      "Epoch 28/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 1.2029 - val_loss: 0.5129\n",
      "Epoch 29/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 1.2721 - val_loss: 0.3627\n",
      "Epoch 30/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 1.1167 - val_loss: 0.4318\n",
      "Epoch 31/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 1.1269 - val_loss: 0.4219\n",
      "Epoch 32/100\n",
      "3/3 [==============================] - 0s 5ms/step - loss: 1.1055 - val_loss: 0.3009\n",
      "Epoch 33/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 1.0503 - val_loss: 0.2991\n",
      "Epoch 34/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 1.0227 - val_loss: 0.2712\n",
      "Epoch 35/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 1.0119 - val_loss: 0.2690\n",
      "Epoch 36/100\n",
      "3/3 [==============================] - 0s 5ms/step - loss: 0.9866 - val_loss: 0.2952\n",
      "Epoch 37/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 0.9896 - val_loss: 0.2361\n",
      "Epoch 38/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 0.9637 - val_loss: 0.2492\n",
      "Epoch 39/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 0.9570 - val_loss: 0.2542\n",
      "Epoch 40/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 0.9685 - val_loss: 0.2441\n",
      "Epoch 41/100\n",
      "3/3 [==============================] - 0s 5ms/step - loss: 0.9479 - val_loss: 0.2093\n",
      "Epoch 42/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 0.9286 - val_loss: 0.2566\n",
      "Epoch 43/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 0.9284 - val_loss: 0.1961\n",
      "Epoch 44/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 0.9115 - val_loss: 0.1992\n",
      "Epoch 45/100\n",
      "3/3 [==============================] - 0s 5ms/step - loss: 0.9106 - val_loss: 0.2034\n",
      "Epoch 46/100\n",
      "3/3 [==============================] - 0s 5ms/step - loss: 0.9256 - val_loss: 0.1984\n",
      "Epoch 47/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 0.9007 - val_loss: 0.1815\n",
      "Epoch 48/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 0.8934 - val_loss: 0.1802\n",
      "Epoch 49/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 0.9582 - val_loss: 0.2174\n",
      "Epoch 50/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 0.9062 - val_loss: 0.2313\n",
      "Epoch 51/100\n",
      "3/3 [==============================] - 0s 5ms/step - loss: 0.9422 - val_loss: 0.3212\n",
      "Epoch 52/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 0.9140 - val_loss: 0.1785\n",
      "Epoch 53/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 0.8821 - val_loss: 0.2081\n",
      "Epoch 54/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 0.9236 - val_loss: 0.2763\n",
      "Epoch 55/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 0.9107 - val_loss: 0.1793\n",
      "Epoch 56/100\n",
      "3/3 [==============================] - 0s 5ms/step - loss: 0.8768 - val_loss: 0.1763\n",
      "Epoch 57/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 0.8791 - val_loss: 0.1770\n",
      "Epoch 58/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 0.8727 - val_loss: 0.1708\n",
      "Epoch 59/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 0.8949 - val_loss: 0.1795\n",
      "Epoch 60/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 0.8727 - val_loss: 0.1791\n",
      "Epoch 61/100\n",
      "3/3 [==============================] - 0s 5ms/step - loss: 0.8792 - val_loss: 0.1705\n",
      "Epoch 62/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 0.9203 - val_loss: 0.1786\n",
      "Epoch 63/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 0.9272 - val_loss: 0.1850\n",
      "Epoch 64/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 0.8790 - val_loss: 0.1907\n",
      "Epoch 65/100\n",
      "3/3 [==============================] - 0s 5ms/step - loss: 0.8703 - val_loss: 0.1751\n",
      "Epoch 66/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 0.8741 - val_loss: 0.1831\n",
      "Epoch 67/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 0.8856 - val_loss: 0.1701\n",
      "Epoch 68/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 0.8900 - val_loss: 0.1722\n",
      "Epoch 69/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 0.8713 - val_loss: 0.1708\n",
      "Epoch 70/100\n",
      "3/3 [==============================] - 0s 5ms/step - loss: 0.8716 - val_loss: 0.1738\n",
      "Epoch 71/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 0.8995 - val_loss: 0.1711\n",
      "Epoch 72/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 0.8653 - val_loss: 0.1724\n",
      "Epoch 73/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 0.8922 - val_loss: 0.2791\n",
      "Epoch 74/100\n",
      "3/3 [==============================] - 0s 5ms/step - loss: 0.9050 - val_loss: 0.1711\n",
      "Epoch 75/100\n",
      "3/3 [==============================] - 0s 12ms/step - loss: 0.8655 - val_loss: 0.1707\n",
      "Epoch 76/100\n",
      "3/3 [==============================] - 0s 12ms/step - loss: 0.8697 - val_loss: 0.2015\n",
      "Epoch 77/100\n",
      "3/3 [==============================] - 0s 5ms/step - loss: 0.9122 - val_loss: 0.1811\n",
      "Epoch 78/100\n",
      "3/3 [==============================] - 0s 5ms/step - loss: 0.8803 - val_loss: 0.2202\n",
      "Epoch 79/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 0.8739 - val_loss: 0.1807\n",
      "Epoch 80/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 0.8753 - val_loss: 0.1716\n",
      "Epoch 81/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 0.8896 - val_loss: 0.1921\n",
      "Epoch 82/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 0.8783 - val_loss: 0.1707\n",
      "Epoch 83/100\n",
      "3/3 [==============================] - 0s 5ms/step - loss: 0.8665 - val_loss: 0.1706\n",
      "Epoch 84/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 0.8823 - val_loss: 0.1816\n",
      "Epoch 85/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 0.8944 - val_loss: 0.1761\n",
      "Epoch 86/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 0.8818 - val_loss: 0.1760\n",
      "Epoch 87/100\n",
      "3/3 [==============================] - 0s 5ms/step - loss: 0.8718 - val_loss: 0.1710\n",
      "Epoch 88/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 0.8824 - val_loss: 0.1765\n",
      "Epoch 89/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 0.8699 - val_loss: 0.1731\n",
      "Epoch 90/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 0.8682 - val_loss: 0.2005\n",
      "Epoch 91/100\n",
      "3/3 [==============================] - 0s 5ms/step - loss: 0.9054 - val_loss: 0.2155\n",
      "Epoch 92/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 0.8795 - val_loss: 0.1791\n",
      "Epoch 93/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 0.8708 - val_loss: 0.1725\n",
      "Epoch 94/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 0.8861 - val_loss: 0.2608\n",
      "Epoch 95/100\n",
      "3/3 [==============================] - 0s 5ms/step - loss: 0.9376 - val_loss: 0.1930\n",
      "Epoch 96/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 0.8790 - val_loss: 0.1859\n",
      "Epoch 97/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 0.8876 - val_loss: 0.2013\n",
      "Epoch 98/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 0.9356 - val_loss: 0.1731\n",
      "Epoch 99/100\n",
      "3/3 [==============================] - 0s 6ms/step - loss: 0.8641 - val_loss: 0.1775\n",
      "Epoch 100/100\n",
      "3/3 [==============================] - 0s 5ms/step - loss: 0.8745 - val_loss: 0.2152\n"
     ]
    }
   ],
   "source": [
    "# Model definition\n",
    "model = tf.keras.Sequential([\n",
    "    tf.keras.layers.Dense(1, input_shape=[1])\n",
    "])\n",
    "\n",
    "model.compile(optimizer='sgd', loss='mean_squared_error')\n",
    "model.summary()\n",
    "\n",
    "# Train the model\n",
    "history = model.fit(x_train, z_train, validation_data=(x_val, z_val), epochs=100, verbose=1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Validate the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1/1 [==============================] - 0s 19ms/step\n",
      "[[ -3.9068348]\n",
      " [ 11.328476 ]\n",
      " [-17.495085 ]\n",
      " [ -4.7303658]\n",
      " [  0.6225817]\n",
      " [ 12.563771 ]\n",
      " [ 15.446128 ]\n",
      " [-10.083313 ]\n",
      " [ 19.56378  ]\n",
      " [ -8.436252 ]]\n",
      "[ -4.7243776   11.64422389 -16.50328585  -2.62246929  -0.70074421\n",
      "  14.05959315  15.10356044 -10.6352968   19.46071391  -8.04358291]\n",
      "Loss: 279.3726451711599\n",
      "R2: -19.08543022439749\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# validate and test the model\n",
    "z_pred = model.predict(x_test)\n",
    "print(z_pred)\n",
    "print(z_test)\n",
    "print('Loss:', np.mean((z_pred - z_test)**2))\n",
    "\n",
    "# R2 score\n",
    "mean_z = np.mean(z_test)\n",
    "ss_tot = np.sum((z_test - mean_z)**2)\n",
    "ss_res = np.sum((z_test - z_pred)**2)\n",
    "r2 = 1 - ss_res/ss_tot\n",
    "print('R2:', r2)\n",
    "\n",
    "# Plot the loss\n",
    "import matplotlib.pyplot as plt\n",
    "plt.plot(history.history['loss'])\n",
    "plt.plot(history.history['val_loss'])\n",
    "plt.title('Model loss')\n",
    "plt.ylabel('Loss')\n",
    "plt.xlabel('Epoch')\n",
    "plt.legend(['Train', 'Validation'], loc='upper right')\n",
    "plt.show()\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Save Frozen Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Frozen model layers: \n",
      "x\n",
      "sequential/dense/MatMul/ReadVariableOp/resource\n",
      "sequential/dense/BiasAdd/ReadVariableOp/resource\n",
      "sequential/dense/MatMul/ReadVariableOp\n",
      "sequential/dense/MatMul\n",
      "sequential/dense/BiasAdd/ReadVariableOp\n",
      "sequential/dense/BiasAdd\n",
      "NoOp\n",
      "Identity\n",
      "Frozen model inputs: \n",
      "[<tf.Tensor 'x:0' shape=(None, 1) dtype=float32>]\n",
      "Frozen model outputs: \n",
      "[<tf.Tensor 'Identity:0' shape=(None, 1) dtype=float32>]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2024-02-10 17:50:40.562310: I tensorflow/core/grappler/devices.cc:75] Number of eligible GPUs (core count >= 8, compute capability >= 0.0): 0 (Note: TensorFlow was not compiled with CUDA or ROCm support)\n",
      "2024-02-10 17:50:40.562658: I tensorflow/core/grappler/clusters/single_machine.cc:361] Starting new session\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "'./frozen_models/linear_model_frozen.pb'"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Convert to a TensorFlow function and get the concrete function\n",
    "full_model = tf.function(lambda x: model(x))\n",
    "full_model = full_model.get_concrete_function(tf.TensorSpec(model.inputs[0].shape, model.inputs[0].dtype))\n",
    "\n",
    "# Get frozen ConcreteFunction\n",
    "frozen_func = convert_variables_to_constants_v2(full_model)\n",
    "frozen_func.graph.as_graph_def()\n",
    "\n",
    "layers = [op.name for op in frozen_func.graph.get_operations()]\n",
    "\n",
    "print(\"Frozen model layers: \")\n",
    "for layer in layers:\n",
    "    print(layer)\n",
    "\n",
    "print(\"Frozen model inputs: \")\n",
    "print(frozen_func.inputs)\n",
    "print(\"Frozen model outputs: \")\n",
    "print(frozen_func.outputs)\n",
    "\n",
    "# Save frozen graph from frozen ConcreteFunction to protobuf\n",
    "tf.io.write_graph(graph_or_graph_def=frozen_func.graph,\n",
    "                  logdir=\"./frozen_models\",\n",
    "                  name=\"linear_model_frozen.pb\",\n",
    "                  as_text=False)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "mlenv",
   "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": "3.10.13"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}