MLP/MLP.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Multilayer Perceptron 多層感知器"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 使用 Scikit Learn 的 MLPClassifier 來辨識數字"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 程式碼參考資料\n",
    "* [Ashing00/Multilayer-Perceptron](https://github.com/Ashing00/Multilayer-Perceptron/blob/master/MLP.py)\n",
    "* [使用sklearn中的神经网络模块MLPClassifier处理分类问题](https://blog.csdn.net/u011311291/article/details/78743393)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Data\n",
    "* [MNIST handwritten digit database, Yann LeCun, Corinna Cortes and Chris Burges](http://yann.lecun.com/exdb/mnist/)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 讀入標準套件"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "%matplotlib inline\n",
    "from sklearn.neural_network import MLPClassifier\n",
    "import numpy as np\n",
    "import os\n",
    "import struct\n",
    "import matplotlib.pyplot as plt\n",
    "import cv2"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 讀取資料的函數"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def load_data(kind='train'):\n",
    "    if kind=='train':\n",
    "        labels_path=os.path.join('train-labels-idx1-ubyte')        \n",
    "        images_path=os.path.join('train-images-idx3-ubyte')\n",
    "    else:\n",
    "        labels_path=os.path.join('t10k-labels-idx1-ubyte')        \n",
    "        images_path=os.path.join('t10k-images-idx3-ubyte')\n",
    "    \n",
    "    with open(labels_path, 'rb') as lbpath:\n",
    "        magic, n = struct.unpack('>II',\n",
    "                                 lbpath.read(8))\n",
    "        labels = np.fromfile(lbpath,\n",
    "                             dtype=np.uint8)\n",
    "\n",
    "    with open(images_path, 'rb') as imgpath:\n",
    "        magic, num, rows, cols = struct.unpack(\">IIII\",\n",
    "                                               imgpath.read(16))\n",
    "        images = np.fromfile(imgpath,\n",
    "                             dtype=np.uint8).reshape(len(labels), 784)\n",
    "\n",
    "    return images, labels"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 讀取資料"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# 讀取 Training Data\n",
    "X_train, y_train = load_data(kind='train')\n",
    "# 讀取 Testing Data\n",
    "X_test, y_test = load_data(kind='t10k')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 顯示 Training Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAWsAAAD8CAYAAACxUoU3AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nUgAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f800e458b70>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# X_train2 為 60000*28*28 的矩陣\n",
    "X_train2=X_train.copy()\n",
    "X_train2=X_train2.reshape(60000,28,28)\n",
    "\n",
    "# 顯示前 100 筆 Training Data 圖形\n",
    "for i in range(100):    \n",
    "    plt.subplot(10,10,i+1),plt.imshow(X_train2[i],cmap = 'gray')\n",
    "    plt.axis('off')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 設置神經網路模型參數"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# hidden_layer_sizes(50,50) 代表有兩層隱藏層，第一層50個神經元，第二層50個神經元\n",
    "mlp = MLPClassifier(solver='sgd', activation='relu',alpha=1e-4,\n",
    "                    hidden_layer_sizes=(50,50), random_state=1,\n",
    "                    max_iter=10,verbose=10,learning_rate_init=.1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 標準化 Training Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# X_train3 為 60000*784 的陣列\n",
    "X_train3 = X_train.reshape(60000,28*28)\n",
    "# 將訓練數據標準化\n",
    "X_train3 = X_train3/255.\n",
    "\n",
    "# Training Data 的第 1 筆數據\n",
    "# X_train3[0]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 訓練模型"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Iteration 1, loss = 0.31433496\n",
      "Iteration 2, loss = 0.13096526\n",
      "Iteration 3, loss = 0.09746443\n",
      "Iteration 4, loss = 0.08289663\n",
      "Iteration 5, loss = 0.06756179\n",
      "Iteration 6, loss = 0.06050956\n",
      "Iteration 7, loss = 0.05404734\n",
      "Iteration 8, loss = 0.04829666\n",
      "Iteration 9, loss = 0.04256343\n",
      "Iteration 10, loss = 0.04180814\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/opt/anaconda/lib/python3.6/site-packages/sklearn/neural_network/multilayer_perceptron.py:564: ConvergenceWarning: Stochastic Optimizer: Maximum iterations (10) reached and the optimization hasn't converged yet.\n",
      "  % self.max_iter, ConvergenceWarning)\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "MLPClassifier(activation='relu', alpha=0.0001, batch_size='auto', beta_1=0.9,\n",
       "       beta_2=0.999, early_stopping=False, epsilon=1e-08,\n",
       "       hidden_layer_sizes=(50, 50), learning_rate='constant',\n",
       "       learning_rate_init=0.1, max_iter=10, momentum=0.9,\n",
       "       nesterovs_momentum=True, power_t=0.5, random_state=1, shuffle=True,\n",
       "       solver='sgd', tol=0.0001, validation_fraction=0.1, verbose=10,\n",
       "       warm_start=False)"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# 訓練模型\n",
    "mlp.fit(X_train3, y_train)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 顯示訓練結果"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training Set Accuracy: 0.988583\n",
      "Test Set Accuracy: 0.963100\n",
      "4\n",
      "softmax\n"
     ]
    }
   ],
   "source": [
    "print (\"Training Set Accuracy: %f\"% mlp.score(X_train3,y_train))\n",
    "print (\"Test Set Accuracy: %f\"% mlp.score(X_test,y_test))\n",
    "print (mlp.n_layers_) #輸入層+隱藏層+輸出層\n",
    "print (mlp.out_activation_)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAWQAAABwCAYAAAApKwoZAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f7ffff4df60>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# 第 1-5 筆 Training Data 測試\n",
    "train_res = mlp.predict(X_train3[:5])\n",
    "# 顯示圖片與預測值\n",
    "for i in range(5):\n",
    "    plt.subplot(2,5,i+1),plt.imshow(X_train2[i],cmap = 'gray')\n",
    "    plt.title(train_res[i],fontsize=40), plt.xticks([]), plt.yticks([])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 預測自己輸入的手寫數字圖"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAJQAAABwCAYAAAD46VlAAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nBXIlVKhArvwfSVc1k8pKEyAAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f7fe8370518>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# num:有幾張圖片\n",
    "num = 2\n",
    "\n",
    "My_X =np.zeros((num,28*28), dtype=int)\n",
    "img_num=[0]*num\n",
    "img_res=[0]*num\n",
    "#輸入手寫數字圖檔28x28=784 pixel，\n",
    "Input_Numer=[0]*num\n",
    "# 圖片檔名\n",
    "Input_Numer[0]=\"0.jpg\"\n",
    "Input_Numer[1]=\"1.jpg\"\n",
    "# Input_Numer[2]=\"2.jpg\"\n",
    "# Input_Numer[3]=\"3.jpg\"\n",
    "# Input_Numer[4]=\"4.jpg\"\n",
    "\n",
    "# 讀取圖片並轉換成陣列\n",
    "for i in range(num):  #read 20 digits picture\n",
    "    img = cv2.imread(Input_Numer[i],0)    #Gray\n",
    "    img_num[i]=img.copy()\n",
    "    img=img.reshape(My_X.shape[1])\n",
    "    My_X[i] =img.copy()\n",
    "    \n",
    "My_X = My_X/255.\n",
    "img_res = mlp.predict(My_X)\n",
    "\n",
    "# 顯示圖片與預測值\n",
    "for i in range(num):\n",
    "    plt.subplot(2,5,i+1),plt.imshow(img_num[i],cmap = 'gray')\n",
    "    plt.title(img_res[i],fontsize=40), plt.xticks([]), plt.yticks([])"
   ]
  }
 ],
 "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": "3.6.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}