180131_AUX_CH3.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "```\n",
    "補助本「ゼロから作るDeep Learning」\n",
    "3章～7章までを輪読会の補足として、読んで行きます。\n",
    "Eli Kaminuma 2018.1.31\n",
    "```"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "### 補助本３章 [2018.1.31]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXcAAAD8CAYAAACMwORRAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n1WIQXfsAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x12f2ea318d0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "#\n",
    "# [3.2.3]STEP関数の視覚化\n",
    "#\n",
    "import numpy as np\n",
    "import matplotlib.pylab as plt\n",
    "\n",
    "\n",
    "def step_function(x):\n",
    "    return np.array(x > 0, dtype=np.int)\n",
    "\n",
    "X = np.arange(-5.0, 5.0, 0.1)\n",
    "Y = step_function(X)\n",
    "plt.plot(X, Y)\n",
    "plt.ylim(-0.1, 1.1)  # 図で描画するy軸の範囲を指定\n",
    "plt.show()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXcAAAD8CAYAAACMwORRAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nVO4iImFI5S4iEob+DxH15vm8DFuTAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x12f38958ef0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# 補助本３章\n",
    "#\n",
    "# [3.2.4]シグモイド関数の視覚化\n",
    "#\n",
    "#import numpy as np\n",
    "#import matplotlib.pylab as plt\n",
    "#-------------------------------\n",
    "#math.expは単一の変数のみ\n",
    "#np.expはベクトルで入力できる\n",
    "#--------------------------------\n",
    "def sigmoid(x):\n",
    "    return 1 / (1 + np.exp(-x))    \n",
    "\n",
    "X = np.arange(-5.0, 5.0, 0.1)\n",
    "Y = sigmoid(X)\n",
    "plt.plot(X, Y)\n",
    "plt.ylim(-0.1, 1.1)\n",
    "plt.show()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXcAAAD8CAYAAACMwORRAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x12f389c2198>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# 補助本３章\n",
    "#\n",
    "# [3.2.5]STEP関数 vs シグモイド関数\n",
    "#\n",
    "def sigmoid(x):\n",
    "    return 1 / (1 + np.exp(-x))    \n",
    "\n",
    "\n",
    "def step_function(x):\n",
    "    return np.array(x > 0, dtype=np.int)\n",
    "\n",
    "x = np.arange(-5.0, 5.0, 0.1)\n",
    "y1 = sigmoid(x)\n",
    "y2 = step_function(x)\n",
    "\n",
    "plt.plot(x, y1)\n",
    "plt.plot(x, y2, 'k--')\n",
    "plt.ylim(-0.1, 1.1) #図で描画するy軸の範囲を指定\n",
    "plt.show()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXYAAAD8CAYAAABjAo9vAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x12f2ea37048>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# 補助本３章\n",
    "#\n",
    "# [3.2.7]　ReLU(ランプ関数、正規化線形関数)\n",
    "\n",
    "def relu(x):\n",
    "    return np.maximum(0, x)\n",
    "\n",
    "x = np.arange(-5.0, 5.0, 0.1)\n",
    "y = relu(x)\n",
    "plt.plot(x, y)\n",
    "plt.ylim(-1.0, 5.5)\n",
    "plt.show()\n",
    "\n",
    "#"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXcAAAD8CAYAAACMwORRAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n6P8D6EP+3znVyLwAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x12f40c6cdd8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "#活性化関数を３種類重ねて表示\n",
    "#\n",
    "#\n",
    "\n",
    "y1 = sigmoid(x)\n",
    "y2 = step_function(x)\n",
    "y3 = relu(x)\n",
    "\n",
    "plt.plot(x, y1)\n",
    "plt.plot(x, y2, 'k--')\n",
    "plt.plot(x, y3, 'g')\n",
    "plt.ylim(-0.1, 1.1) #図で描画するy軸の範囲を指定\n",
    "plt.show()\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 98,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "W1.shape= (2, 3)\n",
      "X.shape= (2,)\n",
      "B1.shape= (3,)\n",
      "A1.shape= (3,)\n",
      "Z1.shape= (3,)\n",
      "A1= [0.3 0.7 1.1]\n",
      "Z1= [0.57444252 0.66818777 0.75026011]\n"
     ]
    }
   ],
   "source": [
    "# [3.4.2]信号伝達の実装　（p.61,図3-18）\n",
    "# 入力層から第一層目の1番目ニューロンへの信号伝達\n",
    "X=np.array([1.0,0.5])\n",
    "W1=np.array([[0.1,0.3,0.5],[0.2,0.4,0.6]])\n",
    "B1=np.array([0.1,0.2,0.3])\n",
    "print('W1.shape=',W1.shape)\n",
    "print('X.shape=',X.shape)\n",
    "print('B1.shape=',B1.shape)\n",
    "\n",
    "A1=np.dot(X,W1)+B1\n",
    "Z1=sigmoid(A1)\n",
    "\n",
    "print('A1.shape=',A1.shape)\n",
    "print('Z1.shape=',Z1.shape)\n",
    "\n",
    "print('A1=',A1)\n",
    "print('Z1=',Z1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 99,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "y= [0.31682708 0.69627909]\n"
     ]
    }
   ],
   "source": [
    "# [3.4.3]信号伝達の実装のまとめ　（p.65）\n",
    "# 入力層から出力層のニューロンへの信号伝達\n",
    "\n",
    "def identity_function(x):\n",
    "    return x\n",
    "\n",
    "def init_network():\n",
    "    network = {}\n",
    "    network['W1'] = np.array([[0.1, 0.3, 0.5], [0.2, 0.4, 0.6]])\n",
    "    network['b1'] = np.array([0.1, 0.2, 0.3])\n",
    "    network['W2'] = np.array([[0.1, 0.4], [0.2, 0.5], [0.3, 0.6]])\n",
    "    network['b2'] = np.array([0.1, 0.2])\n",
    "    network['W3'] = np.array([[0.1, 0.3], [0.2, 0.4]])\n",
    "    network['b3'] = np.array([0.1, 0.2])\n",
    "    return network\n",
    "def forward(network, x):\n",
    "    W1, W2, W3 = network['W1'], network['W2'], network['W3']\n",
    "    b1, b2, b3 = network['b1'], network['b2'], network['b3']\n",
    "    a1 = np.dot(x, W1) + b1\n",
    "    z1 = sigmoid(a1)\n",
    "    a2 = np.dot(z1, W2) + b2\n",
    "    z2 = sigmoid(a2)\n",
    "    a3 = np.dot(z2, W3) + b3\n",
    "    y = identity_function(a3)\n",
    "    return y\n",
    "\n",
    "network = init_network()\n",
    "x = np.array([1.0, 0.5])\n",
    "y = forward(network, x) #入力から出力方向への伝達処理\n",
    "print('y=',y) # [ 0.31682708 0.69627909]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "exp_a= [ 1.34985881 18.17414537 54.59815003]\n",
      "sum_exp_a= 74.1221542101633\n",
      "y= [0.01821127 0.24519181 0.73659691]\n"
     ]
    }
   ],
   "source": [
    "#------------------\n",
    "# [3.5.1]出力層の設計(1)\n",
    "# softmax function (p.67)\n",
    "#-------------------\n",
    "\n",
    "a = np.array([0.3, 2.9, 4.0])\n",
    "\n",
    "exp_a = np.exp(a) # 指数関数\n",
    "print('exp_a=',exp_a)\n",
    "\n",
    "sum_exp_a = np.sum(exp_a) # 指数関数の和\n",
    "print('sum_exp_a=',sum_exp_a)\n",
    "\n",
    "y = exp_a / sum_exp_a\n",
    "print('y=', y)\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 101,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "y(softmax def)= [0.01821127 0.24519181 0.73659691]\n"
     ]
    }
   ],
   "source": [
    "#------------------\n",
    "# [3.5.1]出力層の設計(2)\n",
    "# softmax functionを定義 (p.68)\n",
    "#-------------------\n",
    "def softmax(a):\n",
    "    exp_a = np.exp(a)\n",
    "    sum_exp_a = np.sum(exp_a)\n",
    "    y=exp_a / sum_exp_a\n",
    "    return y\n",
    "\n",
    "a = np.array([0.3, 2.9, 4.0])\n",
    "y = softmax(a)\n",
    "print('y(softmax def)=',y)\n",
    "\n",
    "#縦方向にbroadcastしてしまうので、転置を入れてから\n",
    "#横方向に計算する。\n",
    "\n",
    "#行単位にかけていくように、列から転置している(#多田さん)\n",
    "#speedの問題なので、行列よりベクトルでbroadcastした方が良い(#多田さん)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 110,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "y(softmax def(overflow modified))= [0.01821127 0.24519181 0.73659691]\n"
     ]
    }
   ],
   "source": [
    "#------------------\n",
    "# [3.5.2] ソフトマックス関数の実装上の注意\n",
    "# (p.69)\n",
    "#-------------------\n",
    "def softmax(a):\n",
    "    c = np.max(a) #overflow対策\n",
    "    exp_a = np.exp(a - c)\n",
    "    sum_exp_a = np.sum(exp_a)\n",
    "    y=exp_a / sum_exp_a\n",
    "    return y\n",
    "\n",
    "a = np.array([0.3, 2.9, 4.0])\n",
    "y = softmax(a)\n",
    "print('y(softmax def(overflow modified))=',y)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 113,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "y= [0.01821127 0.24519181 0.73659691]\n",
      "sum(y)= 1.0\n"
     ]
    }
   ],
   "source": [
    "#------------------\n",
    "# [3.5.3] ソフトマックス関数の特徴\n",
    "# (p.70)\n",
    "#-------------------\n",
    "a = np.array([0.3, 2.9, 4.0])\n",
    "y = softmax(a)\n",
    "print('y=',y)\n",
    "print('sum(y)=',np.sum(y))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 102,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(3,)\n",
      "(2,)\n",
      "(2, 1)\n"
     ]
    }
   ],
   "source": [
    "# 計算上注意するbroadcastの特徴（1から8まで解説）\n",
    "# boradcast [1]\n",
    "\n",
    "a=np.array([100, 200, 300])\n",
    "a2=np.array([100,200])\n",
    "a3=np.array([[100],[200]]) #2行1列\n",
    "print(a.shape)\n",
    "print(a2.shape)\n",
    "print(a3.shape)\n",
    "#b=[]\n",
    "#b=np.array([[10 20 30][40 50 60]])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 103,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[100 200 300]\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "array([[101, 202, 303],\n",
       "       [104, 205, 306],\n",
       "       [107, 208, 309]])"
      ]
     },
     "execution_count": 103,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#broadcast (2)\n",
    "a4=np.array([100,200,300]) #1行3列\n",
    "print(a4)\n",
    "b2=np.array([[1,2,3],[4,5,6],[7,8,9]])\n",
    "#縦方向にブロードキャスト\n",
    "a4+b2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 104,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[100 200 300]\n"
     ]
    }
   ],
   "source": [
    "#------------------------\n",
    "# broadcast(3)\n",
    "#-----------------------\n",
    "a5=np.array([100,200,300]) #1行3列→ブロードキャスト時に自動で転置する\n",
    "print(a5)\n",
    "b3=np.array([[1,2,3],[4,5,6]])\n",
    "#縦方向にブロードキャスト\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 105,
   "metadata": {},
   "outputs": [
    {
     "ename": "ValueError",
     "evalue": "shapes (3,) and (2,3) not aligned: 3 (dim 0) != 2 (dim 0)",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mValueError\u001b[0m                                Traceback (most recent call last)",
      "\u001b[1;32m<ipython-input-105-0dc38c01f341>\u001b[0m in \u001b[0;36m<module>\u001b[1;34m()\u001b[0m\n\u001b[0;32m      1\u001b[0m \u001b[1;31m#broadcast(4)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m----> 2\u001b[1;33m \u001b[0ma5\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mdot\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mb3\u001b[0m\u001b[1;33m)\u001b[0m \u001b[1;31m#前からかける、行ベクトルとして扱う\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m      3\u001b[0m \u001b[1;31m#要素数が違うので、エラーになる\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;31mValueError\u001b[0m: shapes (3,) and (2,3) not aligned: 3 (dim 0) != 2 (dim 0)"
     ]
    }
   ],
   "source": [
    "#broadcast(4)\n",
    "a5.dot(b3) #前からかける、行ベクトルとして扱う\n",
    "#要素数が違うので、エラーになる"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 106,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([1400, 3200])"
      ]
     },
     "execution_count": 106,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#broadcast(5)\n",
    "b3.dot(a5) #後ろからかける、列ベクトルとして扱う"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 107,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[ 700 1000]\n",
      "[ 500 1100]\n"
     ]
    }
   ],
   "source": [
    "#broadcast[6]\n",
    "a6=np.array([100,200]) #1次元ベクトル～わかりにくい\n",
    "b4=np.array([[1,2],[3,4]])\n",
    "print(a6.dot(b4))\n",
    "print(b4.dot(a6))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 108,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[ 500],\n",
       "       [1100]])"
      ]
     },
     "execution_count": 108,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#broadcast[7]\n",
    "a7=np.array([[100],[200]]) #2次元行列～明示的にしておく\n",
    "b5=np.array([[1,2],[3,4]])\n",
    "\n",
    "b5.dot(a7)\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 109,
   "metadata": {},
   "outputs": [
    {
     "ename": "ValueError",
     "evalue": "shapes (2,1) and (2,2) not aligned: 1 (dim 1) != 2 (dim 0)",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mValueError\u001b[0m                                Traceback (most recent call last)",
      "\u001b[1;32m<ipython-input-109-1e115627d436>\u001b[0m in \u001b[0;36m<module>\u001b[1;34m()\u001b[0m\n\u001b[0;32m      1\u001b[0m \u001b[1;31m#broadcast[8]\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m----> 2\u001b[1;33m \u001b[0ma7\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mdot\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mb5\u001b[0m\u001b[1;33m)\u001b[0m \u001b[1;31m# 行列数が合わないのでエラー\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[1;31mValueError\u001b[0m: shapes (2,1) and (2,2) not aligned: 1 (dim 1) != 2 (dim 0)"
     ]
    }
   ],
   "source": [
    "#broadcast[8]\n",
    "a7.dot(b5) # 行列数が合わないのでエラー"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 114,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "5\n",
      "(28, 28)\n",
      "(28, 28)\n"
     ]
    }
   ],
   "source": [
    "# 補助本３章\n",
    "#\n",
    "# [3.6.1]　MNISTデータセット\n",
    "# (p.74-75)\n",
    "#\n",
    "from keras.datasets import mnist\n",
    "\n",
    "\n",
    "import sys, os\n",
    "#sys.path.append(os.pardir)  # 親ディレクトリのファイルをインポートするための設定\n",
    "import numpy as np\n",
    "#from dataset.mnist import load_mnist\n",
    "from PIL import Image\n",
    "\n",
    "\n",
    "def img_show(img):\n",
    "    pil_img = Image.fromarray(np.uint8(img))\n",
    "    pil_img.show()\n",
    "\n",
    "    \n",
    "#(x_train, t_train), (x_test, t_test) = load_mnist(flatten=True, normalize=False)\n",
    "(x_train, t_train), (x_test, t_test) = mnist.load_data()\n",
    "\n",
    "img = x_train[0]\n",
    "label = t_train[0]\n",
    "print(label)  # 5\n",
    "\n",
    "print(img.shape)  # (784,)\n",
    "img = img.reshape(28, 28)  # 形状を元の画像サイズに変形\n",
    "print(img.shape)  # (28, 28)\n",
    "\n",
    "img_show(img)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 118,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy:0.9352\n"
     ]
    }
   ],
   "source": [
    "#\n",
    "# [3.6.2]　ニューラルネットワークの推論処理\n",
    "# (p.76-77)\n",
    "#\n",
    "\n",
    "#from keras.datasets import mnist\n",
    "import pickle\n",
    "import sys, os\n",
    "sys.path.append('./')  # 親ディレクトリのファイルをインポートするための設定\n",
    "import numpy as np\n",
    "import pickle\n",
    "from saitobook.dataset.mnist import load_mnist\n",
    "from saitobook.common.functions import sigmoid, softmax\n",
    "\n",
    "\n",
    "def get_data():\n",
    "    #(x_train, t_train), (x_test, t_test) = mnist.load_data()\n",
    "    (x_train, t_train), (x_test, t_test) = load_mnist(normalize=True, flatten=True, one_hot_label=False)\n",
    "    return x_test, t_test\n",
    "\n",
    "\n",
    "def init_network():\n",
    "    with open(\"./sample_weight.pkl\", 'rb') as f:\n",
    "        network = pickle.load(f)\n",
    "    return network\n",
    "\n",
    "\n",
    "def predict(network, x):\n",
    "    W1, W2, W3 = network['W1'], network['W2'], network['W3']\n",
    "    b1, b2, b3 = network['b1'], network['b2'], network['b3']\n",
    "\n",
    "    #print('W1=',W1.shape)\n",
    "    #print('b1=',b1.shape)\n",
    "    a1 = np.dot(x, W1) + b1\n",
    "    z1 = sigmoid(a1)\n",
    "    a2 = np.dot(z1, W2) + b2\n",
    "    z2 = sigmoid(a2)\n",
    "    a3 = np.dot(z2, W3) + b3\n",
    "    y = softmax(a3)\n",
    "\n",
    "    return y\n",
    "\n",
    "\n",
    "x, t = get_data()\n",
    "#print('x=',x.shape)\n",
    "#print('t=',t.shape)\n",
    "network = init_network()\n",
    "accuracy_cnt = 0\n",
    "for i in range(len(x)):\n",
    "    #print('x[i]=',x[i].shape)\n",
    "    y = predict(network, x[i])\n",
    "    p= np.argmax(y) # 最も確率の高い要素のインデックスを取得\n",
    "    if p == t[i]:\n",
    "        accuracy_cnt += 1\n",
    "\n",
    "print(\"Accuracy:\" + str(float(accuracy_cnt) / len(x)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 124,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "x= (10000, 784)\n",
      "x[0]= (784,)\n",
      "w1= (784, 50)\n",
      "w2= (50, 100)\n",
      "w3= (100, 10)\n"
     ]
    }
   ],
   "source": [
    "#\n",
    "# [3.6.3] バッチ処理(1)\n",
    "# (p.80)\n",
    "#　X   W1     W2       W3  -> Y\n",
    "# 784 784x50  50x100 100x10 ->  10\n",
    "\n",
    "\n",
    "x, _ = get_data()\n",
    "network=init_network()\n",
    "w1, w2, w3 = network['W1'], network['W2'], network['W3']\n",
    "print('x=',x.shape)\n",
    "print('x[0]=',x[0].shape)\n",
    "print('w1=',w1.shape)\n",
    "print('w2=',w2.shape)\n",
    "print('w3=',w3.shape)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 127,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy:0.9352\n"
     ]
    }
   ],
   "source": [
    "#\n",
    "# [3.6.3]　バッチ処理(2)\n",
    "# (p.80)\n",
    "#\n",
    "\n",
    "import sys, os\n",
    "sys.path.append('./')  # 親ディレクトリのファイルをインポートするための設定\n",
    "import numpy as np\n",
    "import pickle\n",
    "from saitobook.dataset.mnist import load_mnist\n",
    "from saitobook.common.functions import sigmoid, softmax\n",
    "\n",
    "def get_data():\n",
    "    (x_train, t_train), (x_test, t_test) = load_mnist(normalize=True, flatten=True, one_hot_label=False)\n",
    "    return x_test, t_test\n",
    "\n",
    "\n",
    "def init_network():\n",
    "    with open(\"sample_weight.pkl\", 'rb') as f:\n",
    "        network = pickle.load(f)\n",
    "    return network\n",
    "\n",
    "\n",
    "def predict(network, x):\n",
    "    w1, w2, w3 = network['W1'], network['W2'], network['W3']\n",
    "    b1, b2, b3 = network['b1'], network['b2'], network['b3']\n",
    "\n",
    "    a1 = np.dot(x, w1) + b1\n",
    "    z1 = sigmoid(a1)\n",
    "    a2 = np.dot(z1, w2) + b2\n",
    "    z2 = sigmoid(a2)\n",
    "    a3 = np.dot(z2, w3) + b3\n",
    "    y = softmax(a3)\n",
    "\n",
    "    return y\n",
    "\n",
    "\n",
    "x, t = get_data()\n",
    "network = init_network()\n",
    "\n",
    "batch_size = 100 # バッチの数\n",
    "accuracy_cnt = 0\n",
    "\n",
    "for i in range(0, len(x), batch_size):\n",
    "    x_batch = x[i:i+batch_size]\n",
    "    y_batch = predict(network, x_batch)\n",
    "    p = np.argmax(y_batch, axis=1)\n",
    "    accuracy_cnt += np.sum(p == t[i:i+batch_size])\n",
    "\n",
    "print(\"Accuracy:\" + str(float(accuracy_cnt) / len(x)))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 128,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[0, 1, 2, 3, 4, 5, 6, 7, 8, 9]"
      ]
     },
     "execution_count": 128,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "list(range(0,10))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 129,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[0, 3, 6, 9]"
      ]
     },
     "execution_count": 129,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "list(range(0,10,3))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 130,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "y= [1 2 1 0]\n"
     ]
    }
   ],
   "source": [
    "x = np.array([[0.1, 0.8, 0.1], [0.3, 0.1, 0.6],[0.2, 0.5, 0.3], [0.8, 0.1, 0.1]])\n",
    "y = np.argmax(x, axis=1)#最大値となる要素番号を取得\n",
    "print('y=',y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 134,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(y==t): [ True  True False  True]\n",
      "sum(y==t): 3\n"
     ]
    }
   ],
   "source": [
    "y = np.array([1, 2, 1, 0])\n",
    "t = np.array([1, 2, 0, 0])\n",
    "print('(y==t):',y==t)\n",
    "#[True True False True]\n",
    "print('sum(y==t):',np.sum(y==t))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "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": "3.6.1"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}