linear_regression_map.ipynb

{
 "cells": [
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Data loading"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.datasets import fetch_california_housing\n",
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "california_dataset = fetch_california_housing()\n",
    "x_train, x_test, y_train, y_test = train_test_split(california_dataset.data, california_dataset.target, test_size=0.2, random_state=42)"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Linear regression with MLE\n",
    "## Model training"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "MSE - Train:  1.4565479755401611\n",
      "MSE - Test:  1.4307565689086914\n",
      "Weights:  [[ 1.5577269e-01]\n",
      " [ 1.2762314e-02]\n",
      " [ 2.4280282e-02]\n",
      " [-4.5123370e-03]\n",
      " [ 1.5370875e-05]\n",
      " [-2.9102636e-03]\n",
      " [-3.7413914e-02]\n",
      " [-1.9200975e-02]]\n",
      "Bias:  -0.00015663616\n"
     ]
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "from tqdm import tqdm\n",
    "import torch\n",
    "from linear_regression import LinearRegressionMAP\n",
    "\n",
    "device = torch.device(\"cuda:0\" if torch.cuda.is_available() else \"cpu\")\n",
    "\n",
    "alpha = 100000\n",
    "model = LinearRegressionMAP(torch.Tensor(x_train), torch.Tensor(y_train), alpha, device)\n",
    "model.train()\n",
    "\n",
    "predicted_train = model.predict(torch.Tensor(x_train)).cpu().detach()\n",
    "print(\"MSE - Train: \", torch.mean(torch.square(torch.Tensor(y_train) - predicted_train)).item())\n",
    "\n",
    "predicted_test = model.predict(torch.Tensor(x_test)).cpu().detach()\n",
    "print(\"MSE - Test: \", torch.mean(torch.square(torch.Tensor(y_test) - predicted_test)).item())\n",
    "\n",
    "print(\"Weights: \", model.weights.cpu().numpy())\n",
    "print(\"Bias: \", model.bias.cpu().numpy())"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Model visualization"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot the results\n",
    "from matplotlib.animation import FuncAnimation\n",
    "from IPython import display\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "def plot_predictions(iteration_index, random_indices, var_index, x_test, y_test, weights, bias) -> None:\n",
    "    x_test = x_test[random_indices]\n",
    "    y_test = y_test[random_indices]\n",
    "    y_predicted = (torch.matmul(x_test, weights) + bias).flatten()\n",
    "\n",
    "    # For plotting\n",
    "    x_min = x_test[:,var_index].min() - 1\n",
    "    x_max = x_test[:,var_index].max() + 1\n",
    "    y_min = y_test.min() - 1\n",
    "    y_max = y_test.max() + 1\n",
    "\n",
    "    # Calculate the MSE\n",
    "    mse = torch.mean(torch.square(y_test - y_predicted))\n",
    "\n",
    "    # Clean everything\n",
    "    plt.clf()\n",
    "    plt.xlabel(california_dataset.feature_names[var_index])\n",
    "    plt.ylabel(california_dataset.target_names[0])\n",
    "    plt.xlim(x_min, x_max)\n",
    "    plt.ylim(y_min, y_max)\n",
    "    plt.scatter(x_test[:,var_index].cpu().numpy(), y_test.cpu().numpy()) # Plot test data\n",
    "    plt.scatter(x_test[:,var_index].cpu().numpy(), y_predicted.cpu().numpy(), color=\"k\") # Plot predictions\n",
    "    plt.title(f\"Iteration {iteration_index} - MSE: {mse:.2f}\")\n",
    "\n",
    "# Data to plot\n",
    "x_test = torch.Tensor(x_test)\n",
    "y_test = torch.Tensor(y_test)\n",
    "\n",
    "num_points_to_display = 100\n",
    "variable_index = 0\n",
    "\n",
    "weights = model.weights.cpu().detach()\n",
    "bias = model.bias.cpu().detach()\n",
    "\n",
    "np.random.seed(42)\n",
    "\n",
    "# Select random indices to display over time\n",
    "random_indices = np.random.choice(x_test.shape[0], num_points_to_display)\n",
    "\n",
    "plot_predictions(0, random_indices, variable_index, x_test, y_test, weights, bias)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "lazy-ml",
   "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": "3.11.3"
  },
  "orig_nbformat": 4
 },
 "nbformat": 4,
 "nbformat_minor": 2
}