{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Batch Normalization\n",
    "\n",
    "利点:\n",
    "\n",
    "- 学習を早く進行させることができる\n",
    "- 初期値にそれほど依存しない\n",
    "  - あんまり凝って初期値を設定する必要がない\n",
    "- 過学習を抑制する"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "強制的にノーマライズ(0 ~ 1)を行う。\n",
    "実際には、正規化後に多少値をいじる。それがスケーリング(乗算)とシフト(加算)。\n",
    "\n",
    "このスケーリングとシフトも学習対象で、学習させながら適宜調整する。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "============== 1/16 ==============\n",
      "epoch:0 | 0.097 - 0.108\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\anzuy\\Project\\ManteraStudio-ReadingCircle-DeepLearningFromScratch\\source\\sample\\common\\multi_layer_net_extend.py:101: RuntimeWarning: overflow encountered in square\n",
      "  weight_decay += 0.5 * self.weight_decay_lambda * np.sum(W**2)\n",
      "C:\\Users\\anzuy\\Project\\ManteraStudio-ReadingCircle-DeepLearningFromScratch\\source\\sample\\common\\multi_layer_net_extend.py:101: RuntimeWarning: invalid value encountered in double_scalars\n",
      "  weight_decay += 0.5 * self.weight_decay_lambda * np.sum(W**2)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch:1 | 0.097 - 0.15\n",
      "epoch:2 | 0.097 - 0.174\n",
      "epoch:3 | 0.097 - 0.191\n",
      "epoch:4 | 0.097 - 0.215\n",
      "epoch:5 | 0.097 - 0.233\n",
      "epoch:6 | 0.097 - 0.243\n",
      "epoch:7 | 0.097 - 0.25\n",
      "epoch:8 | 0.097 - 0.273\n",
      "epoch:9 | 0.097 - 0.282\n",
      "epoch:10 | 0.097 - 0.299\n",
      "epoch:11 | 0.097 - 0.299\n",
      "epoch:12 | 0.097 - 0.308\n",
      "epoch:13 | 0.097 - 0.332\n",
      "epoch:14 | 0.097 - 0.344\n",
      "epoch:15 | 0.097 - 0.359\n",
      "epoch:16 | 0.097 - 0.377\n",
      "epoch:17 | 0.097 - 0.381\n",
      "epoch:18 | 0.097 - 0.387\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "No handles with labels found to put in legend.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch:19 | 0.097 - 0.399\n",
      "============== 2/16 ==============\n",
      "epoch:0 | 0.097 - 0.097\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\anzuy\\Project\\ManteraStudio-ReadingCircle-DeepLearningFromScratch\\source\\sample\\common\\multi_layer_net_extend.py:101: RuntimeWarning: overflow encountered in square\n",
      "  weight_decay += 0.5 * self.weight_decay_lambda * np.sum(W**2)\n",
      "C:\\Users\\anzuy\\Project\\ManteraStudio-ReadingCircle-DeepLearningFromScratch\\source\\sample\\common\\multi_layer_net_extend.py:101: RuntimeWarning: invalid value encountered in double_scalars\n",
      "  weight_decay += 0.5 * self.weight_decay_lambda * np.sum(W**2)\n",
      "c:\\users\\anzuy\\project\\manterastudio-readingcircle-deeplearningfromscratch\\.venv\\lib\\site-packages\\numpy\\core\\fromnumeric.py:87: RuntimeWarning: overflow encountered in reduce\n",
      "  return ufunc.reduce(obj, axis, dtype, out, **passkwargs)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch:1 | 0.097 - 0.097\n",
      "epoch:2 | 0.097 - 0.127\n",
      "epoch:3 | 0.097 - 0.18\n",
      "epoch:4 | 0.097 - 0.193\n",
      "epoch:5 | 0.097 - 0.212\n",
      "epoch:6 | 0.097 - 0.236\n",
      "epoch:7 | 0.097 - 0.259\n",
      "epoch:8 | 0.097 - 0.279\n",
      "epoch:9 | 0.097 - 0.296\n",
      "epoch:10 | 0.097 - 0.324\n",
      "epoch:11 | 0.097 - 0.352\n",
      "epoch:12 | 0.097 - 0.365\n",
      "epoch:13 | 0.097 - 0.38\n",
      "epoch:14 | 0.097 - 0.413\n",
      "epoch:15 | 0.097 - 0.431\n",
      "epoch:16 | 0.097 - 0.449\n",
      "epoch:17 | 0.097 - 0.468\n",
      "epoch:18 | 0.097 - 0.477\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "No handles with labels found to put in legend.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch:19 | 0.097 - 0.499\n",
      "============== 3/16 ==============\n",
      "epoch:0 | 0.144 - 0.09\n",
      "epoch:1 | 0.339 - 0.084\n",
      "epoch:2 | 0.503 - 0.121\n",
      "epoch:3 | 0.601 - 0.175\n",
      "epoch:4 | 0.701 - 0.204\n",
      "epoch:5 | 0.754 - 0.242\n",
      "epoch:6 | 0.811 - 0.278\n",
      "epoch:7 | 0.847 - 0.321\n",
      "epoch:8 | 0.882 - 0.345\n",
      "epoch:9 | 0.903 - 0.365\n",
      "epoch:10 | 0.927 - 0.391\n",
      "epoch:11 | 0.942 - 0.41\n",
      "epoch:12 | 0.956 - 0.449\n",
      "epoch:13 | 0.961 - 0.471\n",
      "epoch:14 | 0.968 - 0.491\n",
      "epoch:15 | 0.974 - 0.509\n",
      "epoch:16 | 0.98 - 0.529\n",
      "epoch:17 | 0.986 - 0.548\n",
      "epoch:18 | 0.988 - 0.557\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "No handles with labels found to put in legend.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch:19 | 0.991 - 0.576\n",
      "============== 4/16 ==============\n",
      "epoch:0 | 0.143 - 0.091\n",
      "epoch:1 | 0.318 - 0.115\n",
      "epoch:2 | 0.475 - 0.186\n",
      "epoch:3 | 0.577 - 0.291\n",
      "epoch:4 | 0.601 - 0.339\n",
      "epoch:5 | 0.648 - 0.398\n",
      "epoch:6 | 0.696 - 0.443\n",
      "epoch:7 | 0.703 - 0.501\n",
      "epoch:8 | 0.729 - 0.535\n",
      "epoch:9 | 0.755 - 0.588\n",
      "epoch:10 | 0.758 - 0.619\n",
      "epoch:11 | 0.782 - 0.649\n",
      "epoch:12 | 0.799 - 0.664\n",
      "epoch:13 | 0.816 - 0.696\n",
      "epoch:14 | 0.809 - 0.709\n",
      "epoch:15 | 0.821 - 0.725\n",
      "epoch:16 | 0.828 - 0.746\n",
      "epoch:17 | 0.84 - 0.765\n",
      "epoch:18 | 0.857 - 0.786\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "No handles with labels found to put in legend.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch:19 | 0.857 - 0.794\n",
      "============== 5/16 ==============\n",
      "epoch:0 | 0.089 - 0.113\n",
      "epoch:1 | 0.092 - 0.136\n",
      "epoch:2 | 0.095 - 0.256\n",
      "epoch:3 | 0.096 - 0.377\n",
      "epoch:4 | 0.1 - 0.491\n",
      "epoch:5 | 0.11 - 0.554\n",
      "epoch:6 | 0.111 - 0.623\n",
      "epoch:7 | 0.125 - 0.665\n",
      "epoch:8 | 0.133 - 0.704\n",
      "epoch:9 | 0.148 - 0.726\n",
      "epoch:10 | 0.16 - 0.754\n",
      "epoch:11 | 0.171 - 0.774\n",
      "epoch:12 | 0.178 - 0.797\n",
      "epoch:13 | 0.182 - 0.81\n",
      "epoch:14 | 0.194 - 0.828\n",
      "epoch:15 | 0.199 - 0.834\n",
      "epoch:16 | 0.191 - 0.836\n",
      "epoch:17 | 0.181 - 0.853\n",
      "epoch:18 | 0.177 - 0.864\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "No handles with labels found to put in legend.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch:19 | 0.176 - 0.874\n",
      "============== 6/16 ==============\n",
      "epoch:0 | 0.113 - 0.11\n",
      "epoch:1 | 0.169 - 0.155\n",
      "epoch:2 | 0.118 - 0.326\n",
      "epoch:3 | 0.12 - 0.55\n",
      "epoch:4 | 0.117 - 0.658\n",
      "epoch:5 | 0.127 - 0.717\n",
      "epoch:6 | 0.131 - 0.772\n",
      "epoch:7 | 0.117 - 0.791\n",
      "epoch:8 | 0.117 - 0.807\n",
      "epoch:9 | 0.117 - 0.823\n",
      "epoch:10 | 0.12 - 0.843\n",
      "epoch:11 | 0.117 - 0.866\n",
      "epoch:12 | 0.117 - 0.882\n",
      "epoch:13 | 0.117 - 0.896\n",
      "epoch:14 | 0.117 - 0.911\n",
      "epoch:15 | 0.117 - 0.916\n",
      "epoch:16 | 0.132 - 0.929\n",
      "epoch:17 | 0.158 - 0.938\n",
      "epoch:18 | 0.117 - 0.947\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "No handles with labels found to put in legend.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch:19 | 0.117 - 0.954\n",
      "============== 7/16 ==============\n",
      "epoch:0 | 0.117 - 0.089\n",
      "epoch:1 | 0.117 - 0.355\n",
      "epoch:2 | 0.116 - 0.588\n",
      "epoch:3 | 0.117 - 0.689\n",
      "epoch:4 | 0.117 - 0.754\n",
      "epoch:5 | 0.117 - 0.793\n",
      "epoch:6 | 0.117 - 0.837\n",
      "epoch:7 | 0.117 - 0.858\n",
      "epoch:8 | 0.117 - 0.883\n",
      "epoch:9 | 0.117 - 0.905\n",
      "epoch:10 | 0.116 - 0.921\n",
      "epoch:11 | 0.116 - 0.933\n",
      "epoch:12 | 0.116 - 0.949\n",
      "epoch:13 | 0.116 - 0.956\n",
      "epoch:14 | 0.116 - 0.968\n",
      "epoch:15 | 0.116 - 0.979\n",
      "epoch:16 | 0.116 - 0.981\n",
      "epoch:17 | 0.116 - 0.987\n",
      "epoch:18 | 0.116 - 0.987\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "No handles with labels found to put in legend.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch:19 | 0.117 - 0.989\n",
      "============== 8/16 ==============\n",
      "epoch:0 | 0.116 - 0.115\n",
      "epoch:1 | 0.116 - 0.46\n",
      "epoch:2 | 0.116 - 0.642\n",
      "epoch:3 | 0.116 - 0.763\n",
      "epoch:4 | 0.116 - 0.811\n",
      "epoch:5 | 0.116 - 0.845\n",
      "epoch:6 | 0.116 - 0.885\n",
      "epoch:7 | 0.116 - 0.924\n",
      "epoch:8 | 0.116 - 0.952\n",
      "epoch:9 | 0.116 - 0.965\n",
      "epoch:10 | 0.116 - 0.978\n",
      "epoch:11 | 0.116 - 0.983\n",
      "epoch:12 | 0.116 - 0.991\n",
      "epoch:13 | 0.116 - 0.992\n",
      "epoch:14 | 0.116 - 0.992\n",
      "epoch:15 | 0.116 - 0.995\n",
      "epoch:16 | 0.116 - 0.998\n",
      "epoch:17 | 0.116 - 0.998\n",
      "epoch:18 | 0.116 - 0.999\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "No handles with labels found to put in legend.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch:19 | 0.116 - 0.999\n",
      "============== 9/16 ==============\n",
      "epoch:0 | 0.094 - 0.172\n",
      "epoch:1 | 0.097 - 0.546\n",
      "epoch:2 | 0.099 - 0.648\n",
      "epoch:3 | 0.117 - 0.679\n",
      "epoch:4 | 0.117 - 0.725\n",
      "epoch:5 | 0.117 - 0.8\n",
      "epoch:6 | 0.117 - 0.863\n",
      "epoch:7 | 0.117 - 0.928\n",
      "epoch:8 | 0.117 - 0.965\n",
      "epoch:9 | 0.117 - 0.981\n",
      "epoch:10 | 0.117 - 0.987\n",
      "epoch:11 | 0.117 - 0.992\n",
      "epoch:12 | 0.117 - 0.996\n",
      "epoch:13 | 0.117 - 0.997\n",
      "epoch:14 | 0.117 - 0.996\n",
      "epoch:15 | 0.117 - 0.997\n",
      "epoch:16 | 0.117 - 0.998\n",
      "epoch:17 | 0.117 - 0.999\n",
      "epoch:18 | 0.117 - 0.999\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "No handles with labels found to put in legend.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch:19 | 0.117 - 0.999\n",
      "============== 10/16 ==============\n",
      "epoch:0 | 0.099 - 0.155\n",
      "epoch:1 | 0.116 - 0.648\n",
      "epoch:2 | 0.116 - 0.785\n",
      "epoch:3 | 0.117 - 0.793\n",
      "epoch:4 | 0.117 - 0.811\n",
      "epoch:5 | 0.116 - 0.877\n",
      "epoch:6 | 0.116 - 0.929\n",
      "epoch:7 | 0.116 - 0.968\n",
      "epoch:8 | 0.116 - 0.974\n",
      "epoch:9 | 0.117 - 0.958\n",
      "epoch:10 | 0.117 - 0.99\n",
      "epoch:11 | 0.117 - 0.991\n",
      "epoch:12 | 0.117 - 0.919\n",
      "epoch:13 | 0.117 - 0.992\n",
      "epoch:14 | 0.117 - 0.996\n",
      "epoch:15 | 0.117 - 0.997\n",
      "epoch:16 | 0.117 - 0.997\n",
      "epoch:17 | 0.117 - 0.997\n",
      "epoch:18 | 0.117 - 0.995\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "No handles with labels found to put in legend.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch:19 | 0.117 - 0.997\n",
      "============== 11/16 ==============\n",
      "epoch:0 | 0.117 - 0.099\n",
      "epoch:1 | 0.117 - 0.637\n",
      "epoch:2 | 0.117 - 0.635\n",
      "epoch:3 | 0.117 - 0.596\n",
      "epoch:4 | 0.117 - 0.772\n",
      "epoch:5 | 0.117 - 0.828\n",
      "epoch:6 | 0.117 - 0.85\n",
      "epoch:7 | 0.116 - 0.954\n",
      "epoch:8 | 0.116 - 0.971\n",
      "epoch:9 | 0.117 - 0.981\n",
      "epoch:10 | 0.117 - 0.984\n",
      "epoch:11 | 0.117 - 0.977\n",
      "epoch:12 | 0.116 - 0.99\n",
      "epoch:13 | 0.116 - 0.994\n",
      "epoch:14 | 0.116 - 0.987\n",
      "epoch:15 | 0.116 - 0.995\n",
      "epoch:16 | 0.116 - 0.995\n",
      "epoch:17 | 0.116 - 0.996\n",
      "epoch:18 | 0.116 - 0.991\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "No handles with labels found to put in legend.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch:19 | 0.116 - 0.997\n",
      "============== 12/16 ==============\n",
      "epoch:0 | 0.105 - 0.151\n",
      "epoch:1 | 0.117 - 0.579\n",
      "epoch:2 | 0.117 - 0.644\n",
      "epoch:3 | 0.116 - 0.754\n",
      "epoch:4 | 0.116 - 0.765\n",
      "epoch:5 | 0.116 - 0.648\n",
      "epoch:6 | 0.116 - 0.85\n",
      "epoch:7 | 0.116 - 0.872\n",
      "epoch:8 | 0.116 - 0.875\n",
      "epoch:9 | 0.117 - 0.885\n",
      "epoch:10 | 0.117 - 0.881\n",
      "epoch:11 | 0.116 - 0.973\n",
      "epoch:12 | 0.117 - 0.982\n",
      "epoch:13 | 0.117 - 0.985\n",
      "epoch:14 | 0.117 - 0.988\n",
      "epoch:15 | 0.117 - 0.985\n",
      "epoch:16 | 0.117 - 0.988\n",
      "epoch:17 | 0.117 - 0.987\n",
      "epoch:18 | 0.117 - 0.988\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "No handles with labels found to put in legend.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch:19 | 0.117 - 0.989\n",
      "============== 13/16 ==============\n",
      "epoch:0 | 0.105 - 0.103\n",
      "epoch:1 | 0.117 - 0.435\n",
      "epoch:2 | 0.117 - 0.541\n",
      "epoch:3 | 0.117 - 0.564\n",
      "epoch:4 | 0.117 - 0.664\n",
      "epoch:5 | 0.117 - 0.672\n",
      "epoch:6 | 0.117 - 0.676\n",
      "epoch:7 | 0.117 - 0.657\n",
      "epoch:8 | 0.116 - 0.706\n",
      "epoch:9 | 0.116 - 0.712\n",
      "epoch:10 | 0.116 - 0.709\n",
      "epoch:11 | 0.116 - 0.805\n",
      "epoch:12 | 0.116 - 0.782\n",
      "epoch:13 | 0.116 - 0.803\n",
      "epoch:14 | 0.117 - 0.804\n",
      "epoch:15 | 0.116 - 0.807\n",
      "epoch:16 | 0.117 - 0.811\n",
      "epoch:17 | 0.117 - 0.81\n",
      "epoch:18 | 0.117 - 0.81\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "No handles with labels found to put in legend.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch:19 | 0.117 - 0.805\n",
      "============== 14/16 ==============\n",
      "epoch:0 | 0.117 - 0.118\n",
      "epoch:1 | 0.117 - 0.298\n",
      "epoch:2 | 0.117 - 0.382\n",
      "epoch:3 | 0.117 - 0.484\n",
      "epoch:4 | 0.117 - 0.491\n",
      "epoch:5 | 0.117 - 0.488\n",
      "epoch:6 | 0.117 - 0.452\n",
      "epoch:7 | 0.117 - 0.503\n",
      "epoch:8 | 0.117 - 0.504\n",
      "epoch:9 | 0.117 - 0.493\n",
      "epoch:10 | 0.117 - 0.516\n",
      "epoch:11 | 0.117 - 0.514\n",
      "epoch:12 | 0.117 - 0.512\n",
      "epoch:13 | 0.117 - 0.513\n",
      "epoch:14 | 0.117 - 0.515\n",
      "epoch:15 | 0.117 - 0.572\n",
      "epoch:16 | 0.117 - 0.565\n",
      "epoch:17 | 0.117 - 0.577\n",
      "epoch:18 | 0.117 - 0.586\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "No handles with labels found to put in legend.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch:19 | 0.117 - 0.609\n",
      "============== 15/16 ==============\n",
      "epoch:0 | 0.097 - 0.097\n",
      "epoch:1 | 0.116 - 0.22\n",
      "epoch:2 | 0.116 - 0.383\n",
      "epoch:3 | 0.116 - 0.403\n",
      "epoch:4 | 0.116 - 0.425\n",
      "epoch:5 | 0.116 - 0.412\n",
      "epoch:6 | 0.116 - 0.418\n",
      "epoch:7 | 0.116 - 0.415\n",
      "epoch:8 | 0.116 - 0.429\n",
      "epoch:9 | 0.117 - 0.431\n",
      "epoch:10 | 0.117 - 0.432\n",
      "epoch:11 | 0.117 - 0.426\n",
      "epoch:12 | 0.117 - 0.431\n",
      "epoch:13 | 0.117 - 0.432\n",
      "epoch:14 | 0.117 - 0.431\n",
      "epoch:15 | 0.117 - 0.432\n",
      "epoch:16 | 0.117 - 0.433\n",
      "epoch:17 | 0.117 - 0.524\n",
      "epoch:18 | 0.117 - 0.51\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "No handles with labels found to put in legend.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch:19 | 0.117 - 0.525\n",
      "============== 16/16 ==============\n",
      "epoch:0 | 0.087 - 0.125\n",
      "epoch:1 | 0.117 - 0.257\n",
      "epoch:2 | 0.117 - 0.305\n",
      "epoch:3 | 0.117 - 0.324\n",
      "epoch:4 | 0.117 - 0.325\n",
      "epoch:5 | 0.117 - 0.323\n",
      "epoch:6 | 0.117 - 0.324\n",
      "epoch:7 | 0.117 - 0.396\n",
      "epoch:8 | 0.117 - 0.396\n",
      "epoch:9 | 0.117 - 0.43\n",
      "epoch:10 | 0.117 - 0.43\n",
      "epoch:11 | 0.117 - 0.414\n",
      "epoch:12 | 0.117 - 0.429\n",
      "epoch:13 | 0.117 - 0.42\n",
      "epoch:14 | 0.117 - 0.431\n",
      "epoch:15 | 0.117 - 0.422\n",
      "epoch:16 | 0.117 - 0.428\n",
      "epoch:17 | 0.117 - 0.431\n",
      "epoch:18 | 0.117 - 0.413\n",
      "epoch:19 | 0.117 - 0.43\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 16 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import sys, os\n",
    "sys.path.append(os.path.abspath(os.path.join('..', 'sample')))\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "from dataset.mnist import load_mnist\n",
    "from common.multi_layer_net_extend import MultiLayerNetExtend\n",
    "from common.optimizer import SGD, Adam\n",
    "\n",
    "(x_train, t_train), (x_test, t_test) = load_mnist(normalize=True)\n",
    "\n",
    "# 学習データを削減\n",
    "x_train = x_train[:1000]\n",
    "t_train = t_train[:1000]\n",
    "\n",
    "max_epochs = 20\n",
    "train_size = x_train.shape[0]\n",
    "batch_size = 100\n",
    "learning_rate = 0.01\n",
    "\n",
    "\n",
    "def __train(weight_init_std):\n",
    "    bn_network = MultiLayerNetExtend(input_size=784, hidden_size_list=[100, 100, 100, 100, 100], output_size=10, \n",
    "                                    weight_init_std=weight_init_std, use_batchnorm=True)\n",
    "    network = MultiLayerNetExtend(input_size=784, hidden_size_list=[100, 100, 100, 100, 100], output_size=10,\n",
    "                                weight_init_std=weight_init_std)\n",
    "    optimizer = SGD(lr=learning_rate)\n",
    "    \n",
    "    train_acc_list = []\n",
    "    bn_train_acc_list = []\n",
    "    \n",
    "    iter_per_epoch = max(train_size / batch_size, 1)\n",
    "    epoch_cnt = 0\n",
    "    \n",
    "    for i in range(1000000000):\n",
    "        batch_mask = np.random.choice(train_size, batch_size)\n",
    "        x_batch = x_train[batch_mask]\n",
    "        t_batch = t_train[batch_mask]\n",
    "    \n",
    "        for _network in (bn_network, network):\n",
    "            grads = _network.gradient(x_batch, t_batch)\n",
    "            optimizer.update(_network.params, grads)\n",
    "    \n",
    "        if i % iter_per_epoch == 0:\n",
    "            train_acc = network.accuracy(x_train, t_train)\n",
    "            bn_train_acc = bn_network.accuracy(x_train, t_train)\n",
    "            train_acc_list.append(train_acc)\n",
    "            bn_train_acc_list.append(bn_train_acc)\n",
    "    \n",
    "            print(\"epoch:\" + str(epoch_cnt) + \" | \" + str(train_acc) + \" - \" + str(bn_train_acc))\n",
    "    \n",
    "            epoch_cnt += 1\n",
    "            if epoch_cnt >= max_epochs:\n",
    "                break\n",
    "                \n",
    "    return train_acc_list, bn_train_acc_list\n",
    "\n",
    "\n",
    "# 3.グラフの描画==========\n",
    "weight_scale_list = np.logspace(0, -4, num=16)\n",
    "x = np.arange(max_epochs)\n",
    "\n",
    "for i, w in enumerate(weight_scale_list):\n",
    "    print( \"============== \" + str(i+1) + \"/16\" + \" ==============\")\n",
    "    train_acc_list, bn_train_acc_list = __train(w)\n",
    "    \n",
    "    plt.subplot(4,4,i+1)\n",
    "    plt.title(\"W:\" + str(w))\n",
    "    if i == 15:\n",
    "        plt.plot(x, bn_train_acc_list, label='Batch Normalization', markevery=2)\n",
    "        plt.plot(x, train_acc_list, linestyle = \"--\", label='Normal(without BatchNorm)', markevery=2)\n",
    "    else:\n",
    "        plt.plot(x, bn_train_acc_list, markevery=2)\n",
    "        plt.plot(x, train_acc_list, linestyle=\"--\", markevery=2)\n",
    "\n",
    "    plt.ylim(0, 1.0)\n",
    "    if i % 4:\n",
    "        plt.yticks([])\n",
    "    else:\n",
    "        plt.ylabel(\"accuracy\")\n",
    "    if i < 12:\n",
    "        plt.xticks([])\n",
    "    else:\n",
    "        plt.xlabel(\"epochs\")\n",
    "    plt.legend(loc='lower right')\n",
    "    \n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "一部ではバッチノーマライズをしないほうが良いと出ているが、全体を見るとバッチノーマライゼーションを入れたほうが学習しているのがわかる。"
   ]
  }
 ],
 "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": "3.8.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}