Task_4.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Задание 4. Линейная регрессия"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Данные содержатся в файле `data/lin_reg.txt`. Прочитать их можно следующим так:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "D = np.loadtxt('lin_reg.txt', delimiter=',')\n",
    "X = D[:, :-1]\n",
    "Y = D[:, -1]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "При выполнении заданий **нельзя** пользоваться `scikit-learn` и т.п., только `NumPy`/`JAX`/`TF`/`PyTorch` и `matplotlib`/`seaborn`/`plotly`/etc.\n",
    "\n",
    "1. Нарисуйте график среднеквадратичной ошибки в зависимости от параметра регуляризации $\\alpha$, используя полные данные для обучения и для тестирования. \n",
    "2. Подготовьте исходные данные для 5 fold CV.\n",
    "3. С помощью CV найдите оптимальное значение $\\alpha$ на какой-нибудь сетке значений.\n",
    "4. Постройте валидационную кривую. \n",
    "5. Постройте кривую обучения."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Полные данные, аналитически"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "def train_test_split(X, Y, train_percent : float =0.25):\n",
    "    p = np.random.permutation(len(X))\n",
    "    X1 = X[p]\n",
    "    Y1 = Y[p]\n",
    "    x_train = []\n",
    "    y_train = []\n",
    "    x_test = []\n",
    "    y_test = []\n",
    "    for i in range(len(X)):\n",
    "        if i % 5 == 0:\n",
    "            x_test.append(X1[i])\n",
    "            y_test.append(Y1[i])\n",
    "        else:\n",
    "            x_train.append(X1[i])\n",
    "            y_train.append(Y1[i])\n",
    "    return np.array(x_train), np.array(y_train), np.array(x_test), np.array(y_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "x_train, y_train, x_test, y_test = train_test_split(X, Y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "def Q(X, Y, theta, alpha:float)->float:\n",
    "    l = len(X)\n",
    "    return sum([(Y[i] - X[i] @ theta).T*(Y[i] - X[i] @ theta)  for i in range(l)])/l"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "def find_theta(X, Y, alpha: float)->list:\n",
    "    n = len(X[0])\n",
    "    l = len(X)\n",
    "    theta = np.linalg.inv(X.T @ X + alpha * np.identity(n)) @ X.T @ Y\n",
    "    return theta\n",
    "\n",
    "def solve(x:list, X, Y, alpha:float)->float:\n",
    "    theta = find_theta(X,Y,alpha)\n",
    "    y = np.array(x).T@theta\n",
    "    return y\n",
    "\n",
    "def plot():\n",
    "    alphas = np.arange(0.01, 30, 0.1)\n",
    "    thetas = [find_theta(x_train, y_train, alpha) for alpha in alphas]  # get parameters on train\n",
    "    Qs_train = [Q(x_train, y_train, thetas[i], alphas[i]) for i in range(len(alphas))]  # get loss on train\n",
    "    Qs_test = [Q(x_test, y_test, thetas[i], alphas[i]) for i in range(len(alphas))]  # get loss on test\n",
    "    plt.plot(alphas, Qs_train, label = 'train')\n",
    "    plt.plot(alphas, Qs_test, label = 'test')\n",
    "    plt.title('Среднеквадратичная ошибка')\n",
    "    plt.xlabel('alpha')\n",
    "    plt.ylabel('Q(alpha)')\n",
    "    plt.legend()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 5-fold CV, градиентный спуск"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "def lin_reg_grad_descent(x_train, y_train, x_test, y_test, n_iter, lr, alpha):\n",
    "    l, n = x_train.shape\n",
    "    A = np.hstack([np.ones((l, 1)), x_train])\n",
    "    B = np.hstack([np.ones((x_test.shape[0], 1)), x_test])\n",
    "    theta = np.zeros(n + 1)\n",
    "    losses = []\n",
    "    losses_val = []\n",
    "    for _ in range(n_iter):\n",
    "        y_pred = A.dot(theta) \n",
    "        errors = y_train - y_pred\n",
    "        loss = np.mean(errors**2) + alpha * np.sum(theta**2)\n",
    "        grad = -(2 * A.T @ errors + 2*alpha * theta)/l\n",
    "        theta -= lr * grad\n",
    "        losses.append(loss) \n",
    "        loss_val = sum((y_test-B @ theta)**2)/B.shape[0] + alpha * np.sum(theta**2)\n",
    "        losses_val.append(loss_val)\n",
    "    return loss, loss_val"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "# shuffle arrays\n",
    "def cv_grad_lin_reg(X, Y, n_iter, lr, alpha):\n",
    "    np.random.seed(42)\n",
    "    p = np.random.permutation(len(X))\n",
    "    X1 = X[p]\n",
    "    Y1 = Y[p]\n",
    "    X_folds = [[],[],[],[],[]]\n",
    "    Y_folds = [[],[],[],[],[]]\n",
    "    l = len(X)\n",
    "    for i in range(l):\n",
    "        if i % 5 == 0:\n",
    "            X_folds[0].append(X1[i])\n",
    "            Y_folds[0].append(Y1[i])\n",
    "        elif i % 5 == 1:\n",
    "            X_folds[1].append(X1[i])\n",
    "            Y_folds[1].append(Y1[i])\n",
    "        elif i % 5 == 2:\n",
    "            X_folds[2].append(X1[i])\n",
    "            Y_folds[2].append(Y1[i])\n",
    "        elif i % 5 == 3:\n",
    "            X_folds[3].append(X1[i])\n",
    "            Y_folds[3].append(Y1[i])\n",
    "        elif i % 5 == 4:\n",
    "            X_folds[4].append(X1[i])\n",
    "            Y_folds[4].append(Y1[i])\n",
    "\n",
    "    loss_train_all = []\n",
    "    loss_val_all = []\n",
    "    for i in range(5):\n",
    "        x_train_i = [X_folds[j] for j in range(5) if j != i]\n",
    "        y_train_i = [Y_folds[j] for j in range(5) if j != i]\n",
    "        x_test_i = np.array(X_folds[i])\n",
    "        y_test_i = Y_folds[i]\n",
    "        x_train_i = np.reshape(x_train_i, (int(len(X)*4/5),4))\n",
    "        y_train_i = np.reshape(y_train_i, (int(len(X)*4/5),))\n",
    "\n",
    "        x_train_i = (x_train_i - np.mean(x_train_i, axis=0)) / np.std(x_train_i, axis=0)\n",
    "        x_test_i = (x_test_i - np.mean(x_test_i, axis=0)) / np.std(x_test_i, axis=0)\n",
    "        loss_train, loss_val = lin_reg_grad_descent(x_train_i, y_train_i, x_test_i, y_test_i, n_iter, lr, alpha)\n",
    "        loss_train_all.append(loss_train)\n",
    "        loss_val_all.append(loss_val)\n",
    "    loss_train = np.mean(loss_train_all)\n",
    "    loss_val = np.mean(loss_val_all)\n",
    "    return loss_train, loss_val"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train_loss:  111.03880138265986\n",
      "val_loss:  152.28133975196738\n",
      "alpha_opt:  0.001\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x18fa948d640>"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "alphas = np.arange(0.001, 0.03, 0.001)\n",
    "loss_train_alphas = []\n",
    "loss_val_alphas = []\n",
    "for alpha in alphas:\n",
    "    loss_train, loss_val = cv_grad_lin_reg(X, Y, 250, 1e-2, alpha)\n",
    "    loss_train_alphas.append(loss_train)\n",
    "    loss_val_alphas.append(loss_val)\n",
    "alpha_opt = alphas[np.argmin(loss_val_alphas)]\n",
    "print('train_loss: ', np.min(loss_train_alphas))\n",
    "print('val_loss: ', np.min(loss_val_alphas))\n",
    "print('alpha_opt: ', alpha_opt)\n",
    "\n",
    "plt.plot(alphas, loss_train_alphas, label='train')\n",
    "plt.plot(alphas, loss_val_alphas, label='val')\n",
    "plt.title('Валидационная кривая')\n",
    "plt.xlabel('alpha')\n",
    "plt.ylabel('loss')\n",
    "plt.legend()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x18fa9556d60>"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "loss_train_iter = []\n",
    "loss_val_iter = []\n",
    "iters = range(100, 250)\n",
    "for n_iter in iters:\n",
    "    loss_train, loss_val = cv_grad_lin_reg(X, Y, n_iter, 1e-2, alpha_opt)\n",
    "    loss_train_iter.append(loss_train)\n",
    "    loss_val_iter.append(loss_val)\n",
    "plt.plot(iters, loss_train_iter, label ='train')\n",
    "plt.plot(iters, loss_val_iter, label ='val')\n",
    "plt.xlabel('количество итераций')\n",
    "plt.ylabel('loss')\n",
    "plt.legend()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "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": "3.9.12"
  },
  "vscode": {
   "interpreter": {
    "hash": "916dbcbb3f70747c44a77c7bcd40155683ae19c65e1c03b4aa3499c5328201f1"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}