逻辑回归与小批量梯度下降.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn import datasets\n",
    "from sklearn.model_selection import train_test_split\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "import random\n",
    "\n",
    "x_train, y_train = datasets.load_svmlight_file('a9a')\n",
    "x_test, y_test = datasets.load_svmlight_file('a9a.t',n_features=123)\n",
    "\n",
    "x_train = x_train.toarray()\n",
    "x_test = x_test.toarray()\n",
    "\n",
    "y_train = ( y_train + 1 ) / 2\n",
    "y_test = ( y_test + 1 ) / 2\n",
    "\n",
    "\n",
    "#初始化学习率\n",
    "learnRate = 0.01\n",
    "\n",
    "#计算训练集的行数\n",
    "train_row_size, = y_train.shape\n",
    "test_row_size, = y_test.shape\n",
    "#生成一个行数=train_array_row_size的全 1向量,算出X\n",
    "temp_vector = np.ones(train_row_size)\n",
    "X_train = np.column_stack((x_train,temp_vector))\n",
    "temp_vector = np.ones(test_row_size)\n",
    "X_test = np.column_stack((x_test,temp_vector))\n",
    "batch_size = 30\n",
    "epoch = 8000"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "def computeHypro(x):\n",
    "    return 1 / (1 + np.exp(x))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "def computeGradient_train():\n",
    "    G_array = []\n",
    "    for i in range(batch_size):\n",
    "        temp = random.randint(0,train_row_size-1)\n",
    "        h = 1 / (1 + np.exp(-np.dot(w.T,X_train[temp])))\n",
    "        G = (h - y_train[temp])*X_train[temp]\n",
    "        G_array.append(G)\n",
    "    sum = np.mean(G_array, axis=0)\n",
    "    for n in range (124):        \n",
    "        w[n] = w[n] - learnRate * sum[n]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [],
   "source": [
    "def computeGradient_test():\n",
    "    G_array = []\n",
    "    for i in range(batch_size):\n",
    "        temp = random.randint(0,test_row_size-1)\n",
    "        h = 1 / (1 + np.exp(-np.dot(w.T,X_test[temp])))\n",
    "        G = (h - y_test[temp])*X_test[temp]\n",
    "        G_array.append(G)\n",
    "    sum = np.mean(G_array, axis=0)\n",
    "    for n in range (124):        \n",
    "        w[n] = w[n] - learnRate * sum[n]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "def computeLoss_train():\n",
    "    x = np.dot( -w.T,X_train.T)\n",
    "    h = computeHypro(x)\n",
    "    right_part = np.dot(y_train.T,np.log(h))\n",
    "    \n",
    "    x1 = np.ones(train_row_size) - y_train\n",
    "    y1 = np.log(np.ones(train_row_size) - h)\n",
    "    left_part = np.dot(x1.T,y1)\n",
    "    \n",
    "    loss = -(right_part + left_part) / train_row_size\n",
    "    return loss"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "def computeLoss_test():\n",
    "    x = np.dot( -w.T,X_test.T)\n",
    "    h = computeHypro(x)\n",
    "    right_part = np.dot(y_test.T,np.log(h))\n",
    "    \n",
    "    x1 = np.ones(test_row_size) - y_test\n",
    "    y1 = np.log(np.ones(test_row_size) - h)\n",
    "    left_part = np.dot(x1.T,y1)\n",
    "    \n",
    "    loss = -(right_part + left_part) / test_row_size\n",
    "    return loss"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_loss_tuple = []\n",
    "test_loss_tuple = []\n",
    "#初始化w\n",
    "w = np.zeros(124)\n",
    "for i in range (epoch):\n",
    "    computeGradient_train()\n",
    "    train_loss_tuple.append(computeLoss_train())\n",
    "w = np.zeros(124)\n",
    "for i in range (epoch):\n",
    "    computeGradient_test()\n",
    "    test_loss_tuple.append(computeLoss_test())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.title('Logistic Regression\\nMBGD')\n",
    "plt.xlabel('training time')\n",
    "plt.ylabel('Lvalidation')\n",
    "plt.plot(train_loss_tuple,label= \"Train loss\",color = 'red')\n",
    "plt.plot(test_loss_tuple,label= \"Test loss\",color = 'blue')\n",
    "plt.legend(loc='upper right')\n",
    "plt.grid()\n",
    "plt.show()"
   ]
  }
 ],
 "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": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}