old_Chainer Beginer's Hands-on.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "2017.07.31 Chainer v2向けに変更を行いました。\n",
    "\n",
    "# はじめに\n",
    "\n",
    "[第1回 Chainer Beginner's Hands-on](https://chainer.connpass.com/event/53429/)を大手町にあるPreferred Networksオフィスの多目的室にて開催いたしました。この記事は、このハンズオンで行った内容を記事化したものです。\n",
    "\n",
    "ハンズオン当日に使用した資料等は以下のGithubリポジトリにまとめられています：[mitmul/chainer-handson](https://github.com/mitmul/chainer-handson)\n",
    "\n",
    "## 環境構築について\n",
    "\n",
    "当日は、さくらインターネット様より4つのPascal TITAN Xが搭載されたGPUサーバを20台（合計80GPUs！）、無償にてお借りし、参加者の皆様に使って頂く形でハンズオンを行いました。この場をお借りしてさくらインターネット様に感謝申し上げます。さくら高火力コンピューティングでは、近日、時間単位でのGPUサーバのレンタルサービスが開始されるとのことですので、GPU環境の導入をご検討の方は、ぜひチェックしてみてください。\n",
    "\n",
    "[さくら高火力コンピューティング](https://www.sakura.ad.jp/koukaryoku/)\n",
    "\n",
    "ハンズオン当日はまず、このお借りしたさくら高火力の各ノードにsshログインしていただき、NVIDIA CUDAをインストールするところから始めましたが、この記事ではその部分はスキップし、Chainerの使い方に関する部分からまとめていきます。\n",
    "\n",
    "環境構築の方法については、以下の資料にまとめてありますので、ご参照ください。\n",
    "\n",
    "[さくら高火力コンピューティングサーバ上での環境構築](http://bit.ly/chainer-handson-vol1)\n",
    "\n",
    "これは一部を除けばUbuntu 14.04で動いているNVIDIA GPUが搭載されたサーバに対する環境構築の手順としてお使い頂けます。\n",
    "\n",
    "それでは本題に入っていきます。以下は、Ubuntu14.04に標準でインストールされているPython 3.4を前提として書かれたチュートリアルになっています。あらかじめ上記資料のP.9、P.11あたりを参考に、関連ライブラリとChainer自体のインストールは済ませてください。\n",
    "\n",
    "\n",
    "## 注意\n",
    "\n",
    "**以下のコード部分とそれに続く出力結果は、Jupyter notebook上で実行した場合を想定したものとなっています。**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 学習ループを書いてみよう\n",
    "\n",
    "ここでは、\n",
    "\n",
    "1. データセットからデータを取り出す\n",
    "2. モデルに入力する\n",
    "3. Optimizerを使ってモデルのパラメータを更新して学習を行うループを回す\n",
    "\n",
    "ことをやってみます。これらを通して、基本的な学習ループの書き方が体験できます。（ChainerにはTrainerという便利な学習ループの抽象化クラスが用意されていますが、この章はそれを敢えて用いずに解説したものになります。）\n",
    "\n",
    "## 1. データセットの準備\n",
    "\n",
    "ここでは、Chainerが用意しているMNISTデータセットを使うための便利なメソッドを利用します。これを使うと、データのダウンロードから、一つ一つのデータを取り出せるようにするところまでが隠蔽されます。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAD8CAYAAAC4nHJkAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n4QeCIvxAUIQfCIrwA0H9H4BpmwJXvvG+AAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f16b42ec828>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "label: 5\n"
     ]
    }
   ],
   "source": [
    "from chainer.datasets import mnist\n",
    "\n",
    "# データセットがダウンロード済みでなければ、ダウンロードも行う\n",
    "train, test = mnist.get_mnist(withlabel=True, ndim=1)\n",
    "\n",
    "# matplotlibを使ったグラフ描画結果がnotebook内に表示されるようにします。\n",
    "%matplotlib inline\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "# データの例示\n",
    "x, t = train[0]\n",
    "plt.imshow(x.reshape(28, 28), cmap='gray')\n",
    "plt.show()\n",
    "print('label:', t)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 2. Iteratorの作成\n",
    "\n",
    "データセットから決まった数のデータを取得し、それらを束ねてミニバッチを作成して返してくれる`Iterator`を作成しましょう。これをこの後の学習ループの中で使用していきます。イテレータは、`next()`メソッドで新しいミニバッチを返してくれます。内部ではデータセットを何周なめたか（`epoch`）、現在のイテレーションが新しいepochの最初のイテレーションか、を管理するプロパティ（`is_new_epoch`）などを持っています。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from chainer import iterators\n",
    "\n",
    "batchsize = 128\n",
    "\n",
    "train_iter = iterators.SerialIterator(train, batchsize)\n",
    "test_iter = iterators.SerialIterator(test, batchsize,\n",
    "                                     repeat=False, shuffle=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Iteratorについて\n",
    "\n",
    "- Chainerがいくつか用意しているIteratorの一種である`SerialIterator`は、データセットの中のデータを順番に取り出してくる最もシンプルなIteratorです。\n",
    "- 引数にデータセットオブジェクトと、バッチサイズを取ります。\n",
    "- このとき渡したデータセットオブジェクトから、何周も何周もデータを繰り返し読み出す必要がある場合は`repeat`引数を`True`とし、1周が終わったらそれ以上データを取り出したくない場合はこれを`False`とします。デフォルトでは、`True`になっています。\n",
    "- `shuffle`引数に`True`を渡すと、データセットから取り出されてくるデータの順番をエポックごとにランダムに変更します。\n",
    "\n",
    "ここで、`batchsize = 128`としているので、ここで作成した訓練データ用の`Iterator`である`train_iter`およびテストデータ用の`Iterator`である`test_iter`は、それぞれ128枚の数字画像データを一括りにして返す`Iterator`ということになります。[^TrainingデータとValidationデータ]\n",
    "\n",
    "## 3. モデルの定義\n",
    "\n",
    "ここでは、シンプルな三層パーセプトロンを定義します。これは全結合層のみからなるネットワークです。中間層のユニット数は適当に100とし、出力は10クラスなので10とします。ここで用いるMNISTデータセットは10種のラベルを持つためです。では、モデルを定義するために必要な`Link`, `Function`, そして`Chain`について、簡単にここで説明を行います。\n",
    "\n",
    "### LinkとFunction\n",
    "\n",
    "- Chainerでは、ニューラルネットワークの各層を、`Link`と`Function`に区別します。\n",
    "- **`Link`は、パラメータを持つ関数です。**\n",
    "- **`Function`は、パラメータを持たない関数です。**\n",
    "- これらを組み合わせてモデルを記述します。\n",
    "- パラメータを持つ層は、`chainer.links`モジュール以下にたくさん用意されています。\n",
    "- パラメータを持たない層は、`chainer.functions`モジュール以下にたくさん用意されています。\n",
    "- これらを簡単に使うために、"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import chainer.links as L\n",
    "import chainer.functions as F"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "と別名を与えて、`L.Convolution2D(...)`や`F.relu(...)`のように用いる慣習があります。\n",
    "\n",
    "### Chain\n",
    "\n",
    "- Chainは、パラメータを持つ層＝**`Link`をまとめておくためのクラス**です。\n",
    "- パラメータを持つということは、基本的にモデルの学習の際にそれらを更新していく必要があるということです（例外はあります）。\n",
    "- そこで、学習中に`Optimizer`が更新すべき全てのパラメータを簡単に取得できるように、`Chain`で一箇所にまとめておきます。\n",
    "\n",
    "### Chainを継承して定義されるモデル\n",
    "\n",
    "- モデルは`Chain`クラスを継承したクラスとして定義されることが多いです。\n",
    "- その場合、モデルを表すクラスのコンストラクタで、`with self.init_scope():`のようにして作られたスコープを作り、その中で登録したいリンクオブジェクトをプロパティとして定義しておきます。\n",
    "- また、関数呼び出しのようにしてモデルに`()`アクセサでデータを渡せるように、`__call__`メソッドを定義して、その中にforward処理を記述すると便利です。\n",
    "\n",
    "### GPUで実行するには\n",
    "\n",
    "- `Chain`クラスは`to_gpu`メソッドを持ち、この引数にGPU IDを指定すると、指定したGPU IDのメモリ上にモデルの全パラメータを転送します。\n",
    "- これはモデル内部でのforward/backward計算をその指定したGPU上で行うために必要になります。\n",
    "- これを行わない場合、それらの処理はCPU上で行われます。\n",
    "\n",
    "---\n",
    "\n",
    "それでは、モデルを定義していきます。まずは乱数シードを固定して、本記事とほぼ同様の結果が再現できるようにしておきましょう。（より厳密に計算結果の再現性を保証したい場合は、`deterministic`というオプションについて知る必要があります。こちらの記事が役に立ちます：[ChainerでGPUを使うと毎回結果が変わる理由と対策](http://qiita.com/TokyoMickey/items/cc8cd43545f2656b1cbd)。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import numpy\n",
    "numpy.random.seed(0)\n",
    "import chainer\n",
    "if chainer.cuda.available:\n",
    "    chainer.cuda.cupy.random.seed(0)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "それでは実際にモデルの定義を行い、オブジェクトを作って、GPUに送信してみましょう。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "import chainer\n",
    "import chainer.links as L\n",
    "import chainer.functions as F\n",
    "\n",
    "class MLP(chainer.Chain):\n",
    "\n",
    "    def __init__(self, n_mid_units=100, n_out=10):\n",
    "        # パラメータを持つ層の登録\n",
    "        super(MLP, self).__init__()\n",
    "        with self.init_scope():\n",
    "            self.l1 = L.Linear(None, n_mid_units)\n",
    "            self.l2 = L.Linear(n_mid_units, n_mid_units)\n",
    "            self.l3 = L.Linear(n_mid_units, n_out)\n",
    "\n",
    "    def __call__(self, x):\n",
    "        # データを受け取った際のforward計算を書く\n",
    "        h1 = F.relu(self.l1(x))\n",
    "        h2 = F.relu(self.l2(h1))\n",
    "        return self.l3(h2)\n",
    "\n",
    "gpu_id = 0  # CPUで処理を行いたい場合は、-1をセットしてください。\n",
    "\n",
    "model = MLP()\n",
    "if gpu_id >= 0:\n",
    "    model.to_gpu(gpu_id)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### NOTE\n",
    "\n",
    "ここで、`L.Linear`クラスは全結合層を意味します。コンストラクタの第一引数に`None`を渡すと、実行時に、データがその層に入力された瞬間、必要な数の入力側ユニット数を自動的に計算し、`(n_input)` $\\times$ `n_mid_units`の大きさの行列を作成し、パラメータとして保持します。これは後々、畳み込み層を全結合層の前に配置する際などに便利な機能です。\n",
    "\n",
    "前述のように、`Link`はパラメータを持つので、そのパラメータの値にアクセスすることができます。例えば、上のモデル`MLP`は`l1`という名前の全結合層が登録されています。この全結合相は`W`と`b`という2つのパラメータを持ちます。これらは外からアクセスすることができます。例えば`b`へアクセスするには、以下のようにします。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1つ目の全結合相のバイアスパラメータの形は、 (100,)\n",
      "初期化直後のその値は、 [ 0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.\n",
      "  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.\n",
      "  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.\n",
      "  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.\n",
      "  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.\n",
      "  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.]\n"
     ]
    }
   ],
   "source": [
    "print('1つ目の全結合相のバイアスパラメータの形は、', model.l1.b.shape)\n",
    "print('初期化直後のその値は、', model.l1.b.data)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "ここで、`model.l1.W`を`print`してみると、まだ初期化されておらず何も値が入っていないことがわかります。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "variable W(None)\n"
     ]
    }
   ],
   "source": [
    "print(model.l1.W)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "これは、上記モデルの定義では`Linear`リンクのコンストラクタの第一引数に`None`を渡したため、実行時まで行列`W`は確保されないからです。存在してはいないが、存在する予定であることは`Linear`オブジェクトの内部で把握されています。\n",
    "\n",
    "## 4. 最適化手法の選択\n",
    "\n",
    "Chainerは多くの最適化手法を提供しています。`chainer.optimizers`モジュール以下にそれらはあります。ここでは最もシンプルな勾配降下法の手法である`optimizers.SGD`を用います。Optimizerのオブジェクトには、`setup`メソッドを使ってモデル（`Chain`オブジェクト）を渡します。こうすることでOptimizerは、自身が更新すべきモデル内のパラメータを自動的にたどってくれます。\n",
    "\n",
    "他にもいろいろな最適化手法が手軽に試せるので、色々と試してみて結果の変化を見てみてください。例えば、下の`chainer.optimizers.SGD`のうち`SGD`の部分を`MomentumSGD`, `RMSprop`,  `Adam`などに変えて、結果の違いを見てみると良いでしょう。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "from chainer import optimizers\n",
    "\n",
    "optimizer = optimizers.SGD(lr=0.01)\n",
    "optimizer.setup(model)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### NOTE\n",
    "\n",
    "今回はSGDのコンストラクタの`lr`という引数に $0.01$ を与えました。この値は学習率として知られ、モデルをうまく訓練して良いパフォーマンスを発揮させるために調整する必要がある重要な**ハイパーパラメータ**として知られています。\n",
    "\n",
    "## 5. 学習ループ\n",
    "\n",
    "いよいよ学習ループです。今回は分類問題なので、`softmax_cross_entropy`というロス関数を使って最小化すべきロスの値を計算します。\n",
    "\n",
    "Chainerでは、`Function`や`Link`を使ってモデルのforward計算を行い、結果と正解ラベルを`Function`の一種でありスカラ値を返すロス関数に渡してやり、ロスの計算を行うと、それは他の`Link`や`Function`と同じく、`Variable`オブジェクトを返します。`Variable`オブジェクトはこれまでの計算過程をあとから逆向きに辿り返すための参照を保持しているため、`Variable.backward()`メソッドを呼ぶだけで、自動的にそこからこれまでの計算過程を遡って、途中で施された計算に用いられたパラメータの勾配を計算してくれます。\n",
    "\n",
    "つまり、学習ループの1回の中で行うのは、以下の4項目です。\n",
    "\n",
    "1. モデルにデータを渡して出力`y`を得る\n",
    "2. `y`と正解ラベル`t`を使って、最小化すべきロスの値を`softmax_cross_entropy`関数で計算する\n",
    "3. `softmax_cross_entropy`関数の出力`Variable`の`backward`メソッドを呼んで、モデル内部のパラメータに`grad`プロパティ（これがパラメータ更新に使われる勾配）を持たせる\n",
    "4. Optimizerの`update`メソッドを呼び、3.で計算した`grad`を使って全パラメータを更新する\n",
    "\n",
    "以上です。分類問題ではなく、例えば簡単な回帰問題に取り組むような場合、`F.softmax_cross_entropy`の代わりに`F.mean_squared_error`などを用いることもできます。他にも、いろいろな問題設定に対応するために様々なロス関数がChainerには用意されています。こちらからその一覧を見ることができます：[Loss functions](http://docs.chainer.org/en/latest/reference/functions.html#loss-functions)。\n",
    "\n",
    "では、さっそく訓練ループを書いていきます。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": false,
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch:01 train_loss:0.8068 val_loss:0.7973 val_accuracy:0.8343\n",
      "epoch:02 train_loss:0.4726 val_loss:0.4634 val_accuracy:0.8790\n",
      "epoch:03 train_loss:0.4178 val_loss:0.3770 val_accuracy:0.8981\n",
      "epoch:04 train_loss:0.5055 val_loss:0.3382 val_accuracy:0.9062\n",
      "epoch:05 train_loss:0.2944 val_loss:0.3127 val_accuracy:0.9126\n",
      "epoch:06 train_loss:0.3196 val_loss:0.2960 val_accuracy:0.9158\n",
      "epoch:07 train_loss:0.2762 val_loss:0.2832 val_accuracy:0.9195\n",
      "epoch:08 train_loss:0.1826 val_loss:0.2712 val_accuracy:0.9222\n",
      "epoch:09 train_loss:0.2518 val_loss:0.2612 val_accuracy:0.9251\n",
      "epoch:10 train_loss:0.1884 val_loss:0.2521 val_accuracy:0.9277\n"
     ]
    }
   ],
   "source": [
    "import numpy as np\n",
    "from chainer.dataset import concat_examples\n",
    "from chainer.cuda import to_cpu\n",
    "\n",
    "max_epoch = 10\n",
    "\n",
    "while train_iter.epoch < max_epoch:\n",
    "\n",
    "    # ---------- 学習の1イテレーション ----------\n",
    "    train_batch = train_iter.next()\n",
    "    x, t = concat_examples(train_batch, gpu_id)\n",
    "\n",
    "    # 予測値の計算\n",
    "    y = model(x)\n",
    "\n",
    "    # ロスの計算\n",
    "    loss = F.softmax_cross_entropy(y, t)\n",
    "\n",
    "    # 勾配の計算\n",
    "    model.cleargrads()\n",
    "    loss.backward()\n",
    "\n",
    "    # パラメータの更新\n",
    "    optimizer.update()\n",
    "    # --------------- ここまで ----------------\n",
    "\n",
    "    # 1エポック終了ごとにValidationデータに対する予測精度を測って、\n",
    "    # モデルの汎化性能が向上していることをチェックしよう\n",
    "    if train_iter.is_new_epoch:  # 1 epochが終わったら\n",
    "\n",
    "        # 評価モードにするために、trainコンフィグをFalseにします\n",
    "        with chainer.using_config('train', False):\n",
    "            # ロスの表示\n",
    "            print('epoch:{:02d} train_loss:{:.04f} '.format(\n",
    "                train_iter.epoch, float(to_cpu(loss.data))), end='')\n",
    "\n",
    "            test_losses = []\n",
    "            test_accuracies = []\n",
    "            while True:\n",
    "                test_batch = test_iter.next()\n",
    "                x_test, t_test = concat_examples(test_batch, gpu_id)\n",
    "\n",
    "                # テストデータをforward\n",
    "                y_test = model(x_test)\n",
    "\n",
    "                # ロスを計算\n",
    "                loss_test = F.softmax_cross_entropy(y_test, t_test)\n",
    "                test_losses.append(to_cpu(loss_test.data))\n",
    "\n",
    "                # 精度を計算\n",
    "                accuracy = F.accuracy(y_test, t_test)\n",
    "                accuracy.to_cpu()\n",
    "                test_accuracies.append(accuracy.data)\n",
    "\n",
    "                if test_iter.is_new_epoch:\n",
    "                    test_iter.reset()\n",
    "                    break\n",
    "\n",
    "            print('val_loss:{:.04f} val_accuracy:{:.04f}'.format(\n",
    "                np.mean(test_losses), np.mean(test_accuracies)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "`val_accuracy`に着目してみると、最終的に10エポックで $0.9286$ になっています。おおよそ93%程度の精度で手書きの数字が分類できるようになりました。\n",
    "\n",
    "## 6. 学習済みモデルを保存する\n",
    "\n",
    "Chainerには2つのシリアライズ機能が用意されています。一つはHDF5形式でモデルを保存するもので、もう一つはNumPyのNPZ形式でモデルを保存するものです。今回は、追加ライブラリのインストールが必要なHDF5ではなく、NumPy標準機能で提供されているシリアライズ機能を利用したNPZ形式でのモデルの保存を行います。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "-rw-rw-r-- 1 shunta shunta 333884 Jul 31 23:24 my_mnist.model\r\n"
     ]
    }
   ],
   "source": [
    "from chainer import serializers\n",
    "\n",
    "serializers.save_npz('my_mnist.model', model)\n",
    "\n",
    "# ちゃんと保存されていることを確認\n",
    "%ls -la my_mnist.model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 7. 保存したモデルを読み込んで推論する\n",
    "\n",
    "今しがた保存したNPZファイルを読み込んで、テストデータに対するラベルの予測をネットワークに行わせてみます。NPZファイルにはパラメータが保存されているので、forward計算のロジックを持つモデルのオブジェクトをまず作成し、そのパラメータを先程保存したNPZが持つ値で上書きすることで学習直後のモデルの状態を復元します。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAD8CAYAAAC4nHJkAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nH0iK8ANJEX4gKcIPJEX4gaQIP5DUXwFGhz+pWT5yuQAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f1672a72940>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "label: 7\n"
     ]
    }
   ],
   "source": [
    "# まず同じモデルのオブジェクトを作る\n",
    "model = MLP()\n",
    "\n",
    "# そのオブジェクトに保存済みパラメータをロードする\n",
    "serializers.load_npz('my_mnist.model', model)\n",
    "\n",
    "# GPU上で計算させるために、モデルをGPUに送る\n",
    "if gpu_id >= 0:\n",
    "    model.to_gpu(gpu_id)\n",
    "\n",
    "# テストデータ\n",
    "x, t = test[0]\n",
    "plt.imshow(x.reshape(28, 28), cmap='gray')\n",
    "plt.show()\n",
    "print('label:', t)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "これからモデルに推論させることになるテストデータを表示してみました。以下がこの画像に対し推論を行わせる例になります。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(784,) -> (1, 784)\n",
      "predicted label: 7\n"
     ]
    }
   ],
   "source": [
    "from chainer.cuda import to_gpu\n",
    "\n",
    "# ミニバッチの形にする（ここではサイズ1のミニバッチにするが、\n",
    "# 複数まとめてサイズnのミニバッチにしてまとめて推論することもできる）\n",
    "print(x.shape, end=' -> ')\n",
    "x = x[None, ...]\n",
    "print(x.shape)\n",
    "\n",
    "# GPU上で計算させるため、データもGPU上に送る\n",
    "if gpu_id >= 0:\n",
    "    x = to_gpu(x, gpu_id)\n",
    "\n",
    "# モデルのforward関数に渡す\n",
    "with chainer.using_config('train', False):\n",
    "    y = model(x)\n",
    "\n",
    "# Variable形式で出てくるので中身を取り出す\n",
    "y = y.data\n",
    "\n",
    "# 結果をCPUに送る\n",
    "y = to_cpu(y)\n",
    "\n",
    "# 最大値のインデックスを見る\n",
    "pred_label = y.argmax(axis=1)\n",
    "\n",
    "print('predicted label:', pred_label[0])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Trainerを使ってみよう\n",
    "\n",
    "Trainerを使うと学習ループを陽に書く必要がなくなります。またいろいろな便利なExtentionを使うことで可視化やログの保存などが楽になります。\n",
    "\n",
    "## 1. データセットの準備"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from chainer.datasets import mnist\n",
    "\n",
    "train, test = mnist.get_mnist()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 2. Iteratorの準備"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from chainer import iterators\n",
    "\n",
    "batchsize = 128\n",
    "\n",
    "train_iter = iterators.SerialIterator(train, batchsize)\n",
    "test_iter = iterators.SerialIterator(test, batchsize, False, False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 3. Modelの準備\n",
    "\n",
    "ここでは、先程と同じモデルを再度用いますので、定義は省略します。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "model = MLP()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 4. Updaterの準備\n",
    "\n",
    "Trainerは学習に必要な全てのものをひとまとめにするクラスです。Trainerと、それが内部に持つユーティリティクラスやモデル、データセットクラスなどは、以下のような関係になっています。\n",
    "\n",
    "![image](https://qiita-image-store.s3.amazonaws.com/0/17934/a751df31-b999-f692-d839-488c26b1c48a.png)\n",
    "\n",
    "`Trainer`オブジェクトを作成するときに渡すのは基本的に`Updater`だけですが、`Updater`は中に`Iterator`と`Optimizer`を持っています。`Iterator`からはデータセットにアクセスすることができ、`Optimizer`は中でモデルへの参照を保持しているので、モデルのパラメータを更新することができます。つまり、`Updater`が内部で\n",
    "\n",
    "1. データセットからデータを取り出し（Iterator）\n",
    "2. モデルに渡してロスを計算し（Model = Optimizer.target）\n",
    "3. Optimizerを使ってモデルのパラメータを更新する（Optimizer）\n",
    "\n",
    "という一連の学習の主要部分を行うことができるということです。では、`Updater`オブジェクトを作成してみます。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from chainer import optimizers\n",
    "from chainer import training\n",
    "\n",
    "max_epoch = 20\n",
    "gpu_id = 0  # CPUを用いる場合は、-1をセットしてください\n",
    "\n",
    "# モデルをClassifierで包んで、ロスの計算などをモデルに含める\n",
    "model = L.Classifier(model)\n",
    "if gpu_id >= 0:\n",
    "    model.to_gpu(gpu_id)\n",
    "\n",
    "# 最適化手法の選択\n",
    "optimizer = optimizers.SGD(lr=0.1)\n",
    "optimizer.setup(model)\n",
    "\n",
    "# UpdaterにIteratorとOptimizerを渡す\n",
    "updater = training.StandardUpdater(train_iter, optimizer, device=gpu_id)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### NOTE\n",
    "\n",
    "ここで、上で定義したモデルのオブジェクトを`L.Classifier`に渡して、新しい`Chain`にしています。`L.Classifier`は`Chain`を継承したクラスで、渡された`Chain`を`predictor`というプロパティに保存します。`()`アクセサでデータとラベルを渡すと、中で`__call__`が実行され、まず渡されたデータの方を`predictor`に通し、その出力`y`と、データと一緒に`__call__`に渡されていたラベルを、コンストラクタの`lossfun`引数で指定されたロス関数に渡して、その出力`Variable`を返します。`lossfun`はデフォルトで`softmax_cross_entropy`に指定されています。\n",
    "\n",
    "`StandardUpdater`は前述のようなUpdaterの担当する処理を遂行するための最もシンプルなクラスです。この他にも複数のGPUを用いるための`ParallelUpdater`などが用意されています。\n",
    "\n",
    "## 5. Trainerの設定\n",
    "\n",
    "最後に、`Trainer`の設定を行います。`Trainer`のオブジェクトを作成する際に必須となるのは、先程作成した`Updater`オブジェクトだけですが、二番目の引数`stop_trigger`に学習をどのタイミングで終了するかを表す`(長さ, 単位)`という形のタプルを与えると、指定したタイミングで学習を自動的に終了することができます。長さには任意の整数、単位には`'epoch'`か`'iteration'`のいずれかの文字列を指定できます。`stop_trigger`を指定しない場合、学習は自動的には止まりません。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# TrainerにUpdaterを渡す\n",
    "trainer = training.Trainer(updater, (max_epoch, 'epoch'),\n",
    "                           out='mnist_result')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "`out`引数では、この次に説明する`Extension`を使って、ログファイルやロスの変化の過程を描画したグラフの画像ファイルなどを保存するディレクトリを指定しています。\n",
    "\n",
    "## 6. TrainerにExtensionを追加する\n",
    "\n",
    "`Trainer`を使う利点として、\n",
    "\n",
    "- ログを自動的にファイルに保存（`LogReport`)\n",
    "- ターミナルに定期的にロスなどの情報を表示（`PrintReport`）\n",
    "- ロスを定期的にグラフで可視化して画像として保存（`PlotReport`)\n",
    "- 定期的にモデルやOptimizerの状態を自動シリアライズ（`snapshot`/`snapshot_object`）\n",
    "- 学習の進捗を示すプログレスバーを表示（`ProgressBar`）\n",
    "- モデルの構造をGraphvizのdot形式で保存（`dump_graph`）\n",
    "\n",
    "などなどの様々な便利な機能を簡単に利用することができる点があります。これらの機能を利用するには、`Trainer`オブジェクトに対して`extend`メソッドを使って追加したい`Extension`のオブジェクトを渡してやるだけです。では実際に幾つかの`Extension`を追加してみましょう。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from chainer.training import extensions\n",
    "\n",
    "trainer.extend(extensions.LogReport())\n",
    "trainer.extend(extensions.snapshot(filename='snapshot_epoch-{.updater.epoch}'))\n",
    "trainer.extend(extensions.snapshot_object(model.predictor, filename='model_epoch-{.updater.epoch}'))\n",
    "trainer.extend(extensions.Evaluator(test_iter, model, device=gpu_id))\n",
    "trainer.extend(extensions.PrintReport(['epoch', 'main/loss', 'main/accuracy', 'validation/main/loss', 'validation/main/accuracy', 'elapsed_time']))\n",
    "trainer.extend(extensions.PlotReport(['main/loss', 'validation/main/loss'], x_key='epoch', file_name='loss.png'))\n",
    "trainer.extend(extensions.PlotReport(['main/accuracy', 'validation/main/accuracy'], x_key='epoch', file_name='accuracy.png'))\n",
    "trainer.extend(extensions.dump_graph('main/loss'))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### `LogReport`\n",
    "\n",
    "`epoch`や`iteration`ごとの`loss`, `accuracy`などを自動的に集計し、`Trainer`の`out`引数で指定した出力ディレクトリに`log`というファイル名で保存します。\n",
    "\n",
    "### `snapshot`\n",
    "\n",
    "`Trainer`の`out`引数で指定した出力ディレクトリに`Trainer`オブジェクトを指定されたタイミング（デフォルトでは1エポックごと）に保存します。`Trainer`オブジェクトは上述のように`Updater`を持っており、この中に`Optimizer`とモデルが保持されているため、この`Extension`でスナップショットをとっておけば、学習の復帰や学習済みモデルを使った推論などが学習終了後にも可能になります。\n",
    "\n",
    "### `snapshot_object`\n",
    "\n",
    "しかし、`Trainer`ごと保存した場合、しばしば中身のモデルだけ取り出すのが面倒な場合があります。そこで、`snapshot_object`を使って指定したオブジェクト（ここでは`Classifier`で包まれたモデル）だけを、`Trainer`とは別に保存するようにします。`Classifier`は第1引数に渡された`Chain`オブジェクトを自身の`predictor`というプロパティとして保持してロスの計算を行う`Chain`であり、`Classifier`はそもそもモデル以外にパラメータを持たないので、ここでは後々学習済みモデルを推論に使うことを見越して`model.predictor`を保存対象として指定しています。\n",
    "\n",
    "### `dump_graph`\n",
    "\n",
    "指定された`Variable`オブジェクトから辿れる計算グラフをGraphvizのdot形式で保存します。保存先は`Trainer`の`out`引数で指定した出力ディレクトリです。\n",
    "\n",
    "### `Evaluator`\n",
    "\n",
    "評価用のデータセットの`Iterator`と、学習に使うモデルのオブジェクトを渡しておくことで、学習中のモデルを指定されたタイミングで評価用データセットを用いて評価します。\n",
    "\n",
    "### `PrintReport`\n",
    "\n",
    "`Reporter`によって集計された値を標準出力に出力します。このときどの値を出力するかを、リストの形で与えます。\n",
    "\n",
    "### `PlotReport`\n",
    "\n",
    "引数のリストで指定された値の変遷を`matplotlib`ライブラリを使ってグラフに描画し、出力ディレクトリに`file_name`引数で指定されたファイル名で画像として保存します。\n",
    "\n",
    "---\n",
    "\n",
    "これらの`Extension`は、ここで紹介した以外にも、例えば`trigger`によって個別に作動するタイミングを指定できるなどのいくつかのオプションを持っており、より柔軟に組み合わせることができます。詳しくは公式のドキュメントを見てください：[Trainer extensions](http://docs.chainer.org/en/latest/reference/extensions.html)。\n",
    "\n",
    "## 7. 学習を開始する\n",
    "\n",
    "学習を開始するには、`Trainer`オブジェクトのメソッド`run`を呼ぶだけです。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch       main/loss   main/accuracy  validation/main/loss  validation/main/accuracy  elapsed_time\n",
      "\u001b[J1           0.499081    0.861024       0.252833              0.924545                  1.94617       \n",
      "\u001b[J2           0.23073     0.933202       0.180059              0.947686                  4.33679       \n",
      "\u001b[J3           0.17229     0.95016        0.152644              0.955696                  6.92889       \n",
      "\u001b[J4           0.138831    0.959736       0.120657              0.964794                  9.38878       \n",
      "\u001b[J5           0.116327    0.965952       0.10962               0.967168                  11.8244       \n",
      "\u001b[J6           0.0987802   0.971215       0.100377              0.968849                  14.3496       \n",
      "\u001b[J7           0.0873225   0.974347       0.0907917             0.972706                  16.9415       \n",
      "\u001b[J8           0.0765652   0.977981       0.087784              0.973991                  19.3647       \n",
      "\u001b[J9           0.0684015   0.980011       0.0801619             0.975277                  22.0353       \n",
      "\u001b[J10          0.0611519   0.982193       0.0775525             0.975376                  24.6503       \n",
      "\u001b[J11          0.0550206   0.984375       0.0763309             0.976167                  27.0807       \n",
      "\u001b[J12          0.0495148   0.985777       0.0766076             0.977551                  29.5271       \n",
      "\u001b[J13          0.0459102   0.986624       0.0707709             0.978046                  32.0354       \n",
      "\u001b[J14          0.041151    0.988056       0.0725271             0.978145                  34.7067       \n",
      "\u001b[J15          0.0377337   0.989422       0.0675603             0.979727                  37.2134       \n",
      "\u001b[J16          0.0348712   0.989784       0.0675021             0.978441                  39.7769       \n",
      "\u001b[J17          0.031414    0.991005       0.0719475             0.97854                   42.2335       \n",
      "\u001b[J18          0.0292893   0.991788       0.0731992             0.977453                  44.6412       \n",
      "\u001b[J19          0.0264618   0.992504       0.0694303             0.978343                  47.1684       \n",
      "\u001b[J20          0.0241402   0.993506       0.0717507             0.978145                  49.6553       \n"
     ]
    }
   ],
   "source": [
    "trainer.run()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "初めに取り組んだ学習ループを自分で書いた場合よりもより短いコードで、リッチなログ情報とともに、下記で表示してみるようなグラフなども作りつつ、同様の結果を得ることができました。\n",
    "\n",
    "さっそく保存されているロスのグラフを確認してみましょう。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAg0AAAEKCAYAAACR9WaxAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from IPython.display import Image\n",
    "Image(filename='mnist_result/loss.png')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "精度のグラフも見てみましょう。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAi0AAAEKCAYAAADeqGVhAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAIABJREFUeJzs3XlclXXe//HXWdgFFEFlU0BUENkU3Eolu0ujstG0LDXN\nSW5KUyvvyZlpfpNT09jczZRldw5pOGZqU7m0aaVJWqaImqS4IIJy0BAUEDggcM71++OSIwQIeBAE\nPs/Hg8fZruVzjst5872+i0ZRFAUhhBBCiFuctq0LEEIIIYRoCgktQgghhGgXJLQIIYQQol2Q0CKE\nEEKIdkFCixBCCCHaBQktQgghhGgXGg0ts2fPpkePHgwaNKje1xVFYf78+QQGBhIWFsbBgwctr23b\nto0BAwYQGBjI0qVLW65qIYQQQnQ6jYaWWbNmsW3btgZf37p1K+np6aSnp5OQkMCTTz4JgMlkYu7c\nuWzdupW0tDTWr19PWlpay1UuhBBCiE6l0dAyevRo3NzcGnx9y5YtPPbYY2g0GoYPH05hYSHnz58n\nOTmZwMBAAgICsLW1ZerUqWzZsqVFixdCCCFE56G39gA5OTn4+vpaHvv4+JCTk1Pv8/v27WvSMd3d\n3fHz87O2tJuitLQUJyenti6jQVKfdaQ+60h91rGmvqysLPLz8606f2VlJQaDgfLycquOI8SNsre3\nx8fHBxsbm3pftzq0tJSEhAQSEhIA0Ov1vPbaa21cUf1KSkro0qVLW5fRIKnPOlKfdaQ+61hT36JF\ni6w+v8FgwNnZGT8/PzQajdXHE6I5FEXh4sWLGAwG/P39693G6tDi7e1Ndna25bHBYMDb25vKysp6\nn29IXFwccXFxAERFRRETE2NtaTdFUlLSLVsbSH3WkvqsI/VZp63rKy8vl8Ai2oxGo6F79+7k5eU1\nuI3VQ54nTJjAmjVrUBSFvXv34urqiqenJ9HR0aSnp5OZmUlFRQUbNmxgwoQJ1p5OCCHETSSBRbSl\nxv7+NdrS8sgjj5CUlER+fj4+Pj4sWbKEyspKAOLj44mNjeXLL78kMDAQR0dHEhMT1QPr9Sxfvpxx\n48ZhMpmYPXs2ISEhLfCWhBBCCNEZNRpa1q9ff93XNRoNb7/9dr2vxcbGEhsbe2OVCSGEEM2QkpLC\nmjVrePPNNxvdNj4+nhkzZnDbbbe1QmWipciMuEIIIZptxXcZ7MmoPVppT0Y+K77LaKOK1P6QTQks\nAHv37mX48OE3uaJrqqqqWu1cHZmEFiGEaEO34pd/U4T5uDJv3SFL7Xsy8pm37hBhPq5WHTcrK4ug\noCBmzZpF//79mTZtGtu3b+e2226jX79+JCcnk5yczIgRI4iMjGTkyJGcOHECUDsy33fffQC8+OKL\nzJ49m5iYGAICAmqFmWPHjtG/f390Oh3vvvsu0dHRhIeH8+CDD2I0GgHIzc1l4sSJhIeHEx4ezp49\newBYs2YNYWFhhIeHM2PGDECdhPXjjz+2HL96BFhSUhKjRo1iwoQJDBw4EIDf/OY3DBkyhJCQEMuI\nWVBnkB88eDDh4eHceeedmM1m+vXrZ+mUajabCQwMvG4n1c7glhnyLIQQnVH1l//yRyOBa1/+1Y/b\nypLPjpJ27vJ1t+nhbMdjq5Lp6WJH7uUrBPbowrLt6Szbnl7v9gO9XPjz/Y33bTx16hQfffQR7733\nHtHR0axbt47vv/+eTz/9lFdeeYU1a9awe/du9Ho927dv5w9/+AOffPJJneMcP36cnTt3UlxczIAB\nA3jyySexsbFh69atjB8/HoBJkyYxZ84cAF544QVWrVrF008/zfz58xkzZgybNm3CZDJRUlLC0aNH\nefnll9mzZw/u7u5cunSp0fdy8OBBjhw5YhnC+9577+Hm5kZZWRnR0dE8+OCDmM1m5syZw65du/D3\n9+fSpUtotVqmT5/OBx98wMKFC9m+fTvh4eF4eHg0es6OTEKLEEJYYcV3GYT5uDKyr7vluT0Z+aQa\niogf07fefcxmhcKySvKKr2A2w9RoX3777xT8uijk7lYDS83j3apcHWzo6WJHTmE53l3tcXWof0Kw\n5vL39yc0NBSAkJAQ7rzzTjQaDaGhoWRlZVFUVMTMmTNJT09Ho9FYBof82r333oudnR12dnb06NGD\n3NxcfHx8+OqrryyDRo4cOcILL7xAYWEhJSUljBs3DoBvv/2WNWvWAKDT6XB1dWXNmjVMmTIFd3f1\nz+Z6s8VXGzp0aK05R9588002bdoEQHZ2Nunp6eTl5TF69GjLdtXHnT17Ng888AALFy7kvffe4/HH\nH2/2Z9nRSGgRQnRqNxI6arK0lDwSSZhvV74++gt//vQos2/z5997ssgrvqL+lFyx3M8vuUKVWalz\nrGOXYP7Y3rdEYGlKi0h1q9D8sYGs3XeWBf/Vr0Vqt7Ozs9zXarWWx1qtlqqqKv70pz9xxx13sGnT\nJrKyshqc26bmcXQ6HVVVVRiNRgoLC/Hy8gLUSzubN28mPDyc1atXk5SU1Ox69Xo9ZrMZUC/jVFRU\nWF6rOcNxUlIS27dv58cff8TR0ZGYmJjrzj7s6+tLz549+fbbb0lOTuaDDz5odm0djYQWIUS71mKh\n49FIRgR057sTeSz88Cd+HxvEvtMXKSqrtPxcrnG/5k+VSeHRlbWXKVm2Q71EotNq6O5ki4ezHR7O\ndgT1crbc93C2w6OLHdkFRv76xTFG9VJYu+8sw/t2vyWCy/XUvIw1sq87w/t2r/X4ZioqKrJMVrp6\n9epm7btz507uuOMOy+Pi4mI8PT2prKzkgw8+sBz3zjvv5J133mHhwoWWy0Njx45l4sSJPPvss3Tv\n3p1Lly7h5uaGn58fBw4c4KGHHuLTTz9tsOWnqKiIbt264ejoyPHjx9m7dy8Aw4cP56mnniIzM9Ny\neai6teWJJ55g+vTpzJgxA51O19yPqsOR0CKEaNdqho6Rfd0tX6avPxzOhcvlFBgrKTBWUGisYF92\nJceSMig0VlBoeb4SBxst01buQwNUN4A8/8nPdc6l0YCznR5XRxtcHdSfXq72DPW34cQvxRw8W8hd\nA3syY3gfSyjp5miLTtvwhFl7MvJ55cvjvD1tMBXZR5h656BW+/K3RqqhqFaNI/u6s/zRSFINRTe9\n7t/97nfMnDmTl19+mXvvvbdZ+27dupXJkydbHr/00ksMGzYMDw8Phg0bRnFxMQDLli0jLi6OVatW\nodPpeOeddxgxYgR//OMfGTNmDDqdjsjISFavXs2cOXN44IEHCA8PZ/z48Q2uHzV+/HhWrFhBcHAw\nAwYMsIxe8vDwICEhgUmTJmE2m+nRowfffPMNoE7g+vjjj8uloas0iqLUbaNsY1FRUaSkpLR1GfVq\n62m2GyP1WUfqa76aLR3V9TWnpaO5LSXllSbOF5VzvrCMc1dvD2UXsjs9D1cHGy6VVmCj03Klytzg\nOW31Wro52tDN0RZXB/U2u8DI0XOXGdm3O/eFeVlCSc2fLvb6egNIdVCaPqw3a/edbVbgsPbzq9YS\n/28eO3aM4OBgq45xqxs8eDD79u1rcEG+W01KSgrPPPMMu3fvbutSWs31/h5KS4sQwirWjn6puX+0\nnxtbfz7PHzcfYeaIPqz4LoPzhWXkFJZzvqiM80XlXCqtqHOM7k62dHO05ULxFYJ6OXN7oDtdHW3o\n6qg+r9634fjhA8TeOQZ7G22t6cJ/3Tdj3ljHJocOay+T1BdMRvZ1v6VbWdqzgwcPtnUJTbZ06VLe\neecd6ctSg4QWITo5a/qEKIpCoEcX5t/ZjzlrUvDvopCxI4V7BvVi18l8vj6ay5UqE1cqzZTXc1te\naeZKlQmTWWHau/uo2ey7fKc6T4mzvR4vVwe8utoT7tsVL1d7PF0d8Oxqj5erA71c7Tl4tqBW6Bgb\n3KPeL/28k1ocbGv3C7A2dLTlZRLRsS1evJjFixe3dRm3FAktQnRyDfUJWf5oJIqiUGCsxFBgJPtS\nGdkFRst9Q4ERQ0FZrcswR66otxsP5WCr02Jno8XeRoedvvatvY0WFwcb7G202OnVx2nnLnPYUMTY\noB7MHOmnhpOuDnSxu/5/U20dOqSlRIjWI6FFiE4u3Kcrz48fQNyaA0T4dmV/1iVCvFxY8mkahgIj\npRWmWtt3dbTBp5sD/Xs6MzaoB75ujhSXV5KwK5MRPRX2XdCw/JHB3Nav6V/aezLy+eporqWl5Amd\nhn49nZu0r4QOIToPCS1CtHONXd4xmxVyi8s5e9HImUtGsi8ZOXv1J/uSkfySa31Evj+Vj41Og7HC\nRO/ujowM7I5vN0d8ujng66beOtvX7sBY3dLxznR19MtjdzVv9Iv0CRFCNJWEFiHamLV9Sgb07MJT\naw+yaFx/ejjbs+tkHh8dMBDk6cx/UrIxXCqjwnTtEo5Oq8Grqz293Ry5a2BPfN0cKa80sfqHLB6O\n9uWTgzn8v/sHNvlLv2ZLR1J281s6pE+IEKKpJLQI0cZ+PXrmm7Rf+P3Gn5k3NpAvUs9zyVjBpZIK\nCowVHM8sZ1XGPi5efXyptMLSp+SFzUctx3Sw0WIyKwT1cuaugT3p7eZo+fHq6oCN7tpaqdUtHStm\nDGFkX3fuCOrRqi0d0lIiblSXLl0oKSnh3LlzzJ8/v9aihdViYmJ47bXXiIqKavA4b7zxBnFxcTg6\nOgIQGxvLunXr6Nq16w3VtXfvXlatWsW77757Q/tXW7FiBY6Ojjz22GONbjtkyBD27NlTaxbgjkhC\nixBWak5LSaXJzPnCcrILrl2eOXvJSDdHmzqjZ/76xfFa+zrb63HUmvGyqcLT1Z6BXi7qUF8nW9wc\nbdmVnsfnqeeJG+XP72ODaw3pvR5p6RA35Ps3wHsw+I++9lzmLsg5CLcvbNVSvLy86g0sTfXGG28w\nffp0S2j58ssvraqn5oKM1oiPj2/SdpmZmXh7e7dqYKmqqkKvb/0IIaFFCCv9ehr4bUd+4flPUnls\nRB+Wf5tO9qUyNaAUGDlfVI6pxpozeq0G724O+HZzxN5Gx9Fzl4kZ4MGUIb64Odni5mRLNyd18jMb\nnfbq5GO31alhT0Y+ezIuWjqyxgTVP+S3PtLSIW6I92D4aBZMWa0Gl8xd1x7foMWLF+Pr68vcuXMB\nePHFF9Hr9ezcuZOCggIqKyt5+eWXeeCBB2rtl5WVxX333ceRI0coKyvj8ccf5/DhwwQFBVFWVmbZ\n7sknn2T//v2UlZUxefJklixZwptvvsm5c+e44447cHd3Z+fOnfj5+ZGSkoK7uzv//Oc/ee+99wB1\nSv2FCxeSlZXFPffcw+23386ePXvw9vZmy5YtODg4ALBjxw6effZZVq9ezebNmyktLSU9PZ1FixZR\nUVHB+++/j52dHV9++SVubm68++67JCQkUFFRQWBgIO+//z6Ojo68+OKLdOnShUWLFhETE8OwYcPY\nuXMnhYWFrFq1ilGjRgGwbds2S0iq7z0C7N+/nwULFlBaWoqdnR07duzA0dGR559/nm3btqHVapkz\nZw5PP/10rfefkpLCokWLSEpK4sUXXyQjI4PTp0/Tu3dv/va3vzFjxgxKS0sBWL58OSNHjgTg1Vdf\nZe3atWi1Wu655x7mzJnDlClTLPPkpKen8/DDDzd73hwJLULcoEqTmRO/FHPmopEI367MWJWMBiwL\n4VXPM+LhbIdvNweG9OlGbzdHfLs54uvmiK+bA71c7NHrtHUmN4sbHcCIvt2bVEdbrgEjOrCti+GX\nuksZ1OLsCe9PVG+Lz4NHECS9qv7Up1co3LO0wcM9/PDDLFy40BJa/vOf//DVV18xf/58XFxcyM/P\nZ/jw4UyYMKHBlsR33nkHR0dHjh07RmpqKoMHD7a89te//hU3NzdMJhN33nknqampzJ8/n3/+85/s\n3LnTsnpztQMHDpCYmMi+fftQFIVhw4YxZswYunXrRnp6OuvXr+fdd9/loYce4pNPPmH69Onk5+dj\nY2ODq6sroK4ifejQIcrLywkMDOTVV1/l0KFDPPPMM6xZs4aFCxcyadIk5syZA8ALL7zAqlWrePrp\np+u8t6qqKpKTk/nyyy9ZsmQJ27dvB9TQ8vrrrzf4HoOCgnj44Yf58MMPiY6O5vLlyzg4OJCQkEBW\nVhY//fQTer2eS5cuNfhnUy0tLY3vv/8eBwcHjEYj33zzDfb29qSnp/PII4+QkpLC1q1b2bJlC/v2\n7cPR0dGylpKrqys//fQTERERJCYm3tDSBBJaRKfXlMs7VSYzGXmlHDYU8rOhiNScIo6dv0zF1f4k\nLvZ6vFztyS4oY3Q/d2aO9KO3myM+3RzrTGb2a209z4gQN8y+qxpYirLB1Vd9bIXIyEguXLjAuXPn\nyMvLo1u3bvTq1YtnnnmGXbt2odVqycnJITc3l169etV7jF27djF//nwAwsLCCAsLs7z2n//8h4SE\nBKqqqjh//jxpaWm1Xv+177//nokTJ1rWEpo0aRK7d+9mwoQJ+Pv7ExERAaj9SbKysgD4+uuvufvu\nuy3HuOOOO3B2dsbZ2RlXV1fuv/9+AEJDQ0lNTQXUYPPCCy9QWFhISUkJ48aNq7eeSZMm1TlfRUUF\nBoOBgICABt+jRqPB09OT6OhoAFxcXADYvn078fHxlss81Ys0Xs+ECRMsLUqVlZXMmzePn376CZ1O\nx8mTJy3Hffzxxy2X22ou/piYmMg///lPPvzwQ5KTkxs9369JaBGd3q+nof8hPZ+n1h1k2rDeLPns\nKD8bijh67jJllep8JV3s9AzydmHWSD9CvV0J83Elp6CMeeuvtZQ42OpknhHRvl2nRcSi+pLQ6N9B\nyiqIeb52H5cbMGXKFD7++GN++eUXHn74YT744APy8vI4cOAANjY2+Pn5UV5e3uzjZmZm8tprr7F/\n/366devGrFmzbug41Wr2H9HpdJbLUFu3buXZZ5+tdzutVmt5rNVqqaqqAmDWrFls3ryZ8PBwVq9e\nTVJS0nXPqdPpLPvu3r2b22+/vUXfo16vx2xWfyH79f41F4N8/fXX6dmzJ4cPH8ZsNmNvb3/d4z74\n4IMsWbKEsWPHMmTIELp3b1prck3axjcRouNSFAUvVwceHdab2Yn7eS6plGmr9lFUVsn/JWWwITkb\njQYeGdqbNx6OYMdzY0j9891siBvBH2KDuT/ci5xCNbAsfzSSZ+8ewPJHI5m37hB7MvKbVEP8mL51\nAsbIvu7NWixPiFZXsw/L2D+qtx/NUp+3wsMPP8yGDRv4+OOPmTJlCkVFRfTo0QMbGxt27tzJmTNn\nrrv/6NGjWbduHaC2YFS3Zly+fBknJydcXV3Jzc1l69atln2cnZ0tqzvXNGrUKDZv3ozRaKS0tJRN\nmzZZ+pHUR1EUUlNTLS0wTVVcXIynpyeVlZXNXmdo27Zt3HPPPUDD73HAgAGcP3+e/fv3W85XVVXF\nXXfdxb/+9S9LAKq+POTn58eBAwcA+OSTTxo8d1FREZ6enmi1Wt5//31MJvUXu7vuuovExESMRmOt\n49rb2zNu3DiefPLJG161ukktLdu2bWPBggWYTCaeeOKJOmshFBQUMHv2bDIyMrC3t+e9995j0KBB\ngJrEVq5ciUajITQ0lMTExEbTmBDN0dTRO4qiYCgo4+ecIlINRfyco17quVyu/oPVaeBiFYR5u/LY\nSD/CfFzp69Gl3lV9a5LLM6JTyjl4rRMuqLdTVqvPW9HaEhISQnFxMd7e3nh6ejJt2jTuv/9+QkND\niYqKIigo6Lr7V38hBgcHExwczJAhQwAIDw8nMjKSoKAgfH19ue22ax3a4+LiGD9+PF5eXuzcudPy\n/ODBg5k1axZDhw4F1MsbkZGRlkszv3bgwAEiIyObPHKv2ksvvcSwYcPw8PBg2LBh9QaohiQlJfGX\nv/zluu/R1taWDz/8kKeffpqysjIcHBzYvn07TzzxBCdPniQsLAwbGxvmzJnDvHnz+POf/8xvf/tb\n/vSnP113VfmnnnqKBx98kDVr1jB+/HhLK8z48eP56aefiIqKwtbWltjYWF555RUApk2bxqZNm2pd\nQmsWpRFVVVVKQECAkpGRoVy5ckUJCwtTjh49WmubRYsWKS+++KKiKIpy7NgxZezYsYqiKIrBYFD8\n/PwUo9GoKIqiTJkyRUlMTGzslMqQIUMa3aat7Ny5s61LuK7OWN8Pp/KUyL98rfxwKq/W488O5yjb\njpxX/nfbcWXGqn1KxJKvlD7Pf670ef5zJfAPXyj3v7Vb+cPGVGVD8hll3d4zSsSSr5SnE76qdaxb\nzS3557v7dUU5/Z2iKDXqO/2d+vwt5pb8/Gqwpr6W+H8zLS3N6mN0Zi+99JKyfv36Vjtfdna2Mn78\n+FY7X0v43//9X+WFF1647jbX+3vYaEtLcnIygYGBlk4+U6dOZcuWLQwcONCyTVpamqX1JSgoiKys\nLHJzcwG1t3NZWRk2NjYYjUa8vLxuLF0J0YARAd1ZMiGEuDUHGOjpwqHsAhxs9MxbdwhQZ4Dt39OZ\nuwf2ItRH7YMyoJczdnq1g2x1R9i3p6nT0E8d27xp6Du9mkNfoUWGvgrRHr3wwgutej4fH59al7lu\ndRMnTiQjI4Nvv/32ho/RaGjJycnB19fX8tjHx4d9+/bV2iY8PJyNGzcyatQokpOTOXPmDAaDgSFD\nhrBo0SJ69+6Ng4MDd999d4NNQgkJCSQkJABgMBga7IjU1kpKSm7Z2qB91vfl6Qr8XXUEd782yubY\nRROZRSZiA2xrbWsyK/xSqnCm2MzZyybOFps5e9lMSaX6enLWJZxtINRNwd/VFj9XLb2dtdjqTMAl\nKL/EpVPw46na558zUEdF9hFKSkrokn2EOQO1bP7uIBXZtc/f1ur7/HzPbqTYOZDCbtdGQXQtSMW5\n+BTZvSc1eszm7q8xm7C7cgFH4zkcys7haDyHs7Yrzv+ewO1aW8xKFUUuQRR/u4ordp9Rbu9OuX0P\nrth5UKV3gl81nbdm/Tfj30dL1l9dX3P2F6K92LRpk9XHaJHRQ4sXL2bBggVEREQQGhpKZGQkOp2O\ngoICtmzZQmZmJl27dmXKlCmsXbuW6dOn1zlGXFwccXFxAERFRV33OlpbUif3imnrMhrUHuuz9a0e\n8hvOyL7u7MnI593dh/jHlHCcHWxJO3+ZtHNFpJ27zPFfii3T1tvqtQT1cua+ABfsbXRsPGjg0WG9\n+U+Kgbn3Nr2VpGY51fXFNLTxzdCMmUXr/fPN1Nad5Cv5DZiymr5N6VvQ0P6x/6CvkxYunoKLGVd/\nTkFBFpgrr+1v5wrd+4JjMPoLadC1D910V+h2fhtU/Wrkgq0zdPVVh8e6+qj3vXvA/tfgvmUwYDxk\n7oZ9r8MDb9O3dxgoCigmUMxgvnqrmK8+p4DrcPjq93D3y9BnJFw83eD7r/fzs3Zm1xb8/JPOdCGm\nj7Z5+7cwRVGa3SdDiJaiKMp1X280tHh7e5OdnW15bDAY8Pb2rrWNi4sLiYmJlhP6+/sTEBDAV199\nhb+/Px4eHoA6xnzPnj31hhbReY3s685bUyP57/cPEOrtSkpWAd272DL73ylU//11dbAhxMuFGcP7\nMNDLhRAvVwI8nLCpMTFb9do5o/t7tK/LO02ZWdRshspSbK8UqOGholT9qbx6GzkD1k8Fn6GQvQ9C\np4Bhv3rfXPNL3vyrL/+rgaD3CPhgCnTtowYTrR4+nnXt/Hp7cOsLPYIh+D7oHnjtx7E7ZO2Gj2aR\n1ech/PJ2qLX7jYLSPHUOj8Js9bbIcPX+WTAkQ1nBtXPUPB/Ahkea9zl+Ou/afRsn+GIROPdS5xG5\neutx4SKcdbj6uBfo7Rr//M0mMF5S34sxH0rzwXhRfVyarz7n7AlrfgO2TlBRoj7e+rz6Oeps1Fut\nDeiu3mr11+7rbMAzEj6YTKSjPyTn1e7g2ors7e25ePEi3bt3l+AiWp2iKFy8ePG6g3UaDS3R0dGk\np6db1jbYsGGDZThZtcLCQhwdHbG1tWXlypWMHj0aFxcXevfuzd69ezEajTg4OLBjx47rLlolOhez\nWeFQdgGfHT7P1iPnKS6vYk/GRZzt9IR6u/JwtC8DPV0I8XbFy9W+wf9E23z0zo38pm6qvPoFfgYK\nzkDfsbD2QfXL7nIOuPjAlnlQabwaTtShgyMBfrxOLaevjnw4+O96XtSARgtanXqrqb7Vglarvp5/\nAtwCoP94tfWke6AaVly8r25Tjxpf8llnzPjFzKgdArr0AO8h9e97pRiKctRAs+9fcOob8B8D/e6q\nXaNWe63WWnXXuH90MxzbAn6j1XBVfB5KcuHsHij+BUwVhACk/f3a+R26qZ95196wdrK634U0cO8H\nXzynhpKyAqCB3/4cuoGTBzi6Q/cAyE9Xj+E+AMxV6o+pUm2ZMlVBhfHafXPV1fuV6n2NHtfLx9U5\nT9ogsIB6+d9gMJCXl9cm5xfC3t4eHx+fBl9vNLTo9XqWL1/OuHHjMJlMzJ49m5CQEFasWAGoCzod\nO3aMmTNnotFoCAkJYdWqVQAMGzaMyZMnM3jwYPR6PZGRkZZLQKJzUhSFn7IL+fzwOb78+Tznisqx\n1WsJ83al5IqJaUN9+fhgDrNu82v62jm6z0A7GLj2H/1IbRojdQeBJjTvW3t5oKHf1O/9J5zdq4aS\n6nBSfXs5R23hqKbRqb+lF55RWzt6DVIf2zqBjSPYdgFbJ05m5dB/UOS112y7qK/npsG230HkY3Do\nfZiYAAGja3zRa+v0JamluuaRT6uThA24p+lfnDWHvp5Jat7QVztn6BEEpRfg3MFrk5SNXtS8L+7M\nXXDm+2v7j/mf2vsrCpQVsP/bT4kO8lFDTPH5q7dX7+ts4fxP4NBdndnVyV0NI5bb7tcCipM7OLip\nrSU1P7/q89/z9+bX/9EssrzuxS9lFfiPapPgYmNjg7+/f6ufV4im0iiNXUBqA1FRUaSkpLR1GfVq\nj31G2pqiKBw9d5nPUs/xSXIm+WUKNjoNo/t5cF+4Jy72NvzPx6mW1pJfT2vfqJrN+b9u3m9Sn4Jd\nNfoUmNU+BfXtb6pUf+s2XoKyS1dvC9T75w7Dic/B2QsKz6ohoWa/D4AuPdVA0q1P3dtLmfDJbyHq\nt+qXXgO+ePkeAAAgAElEQVS119+npeXe/w3t31h9N/v8zdi/wfqq92nk87/Z9V/3718jbuX/N4Vo\nKTKNv7BavZO7ncpn+7FcHGx1fJF6nqyLRvRaDcFuWp6/N4S7Q3rh6mADwN41f2LN2KEMqnF5Z83Y\nckp+WAZ9X2q8AN9hMO4V2DBNvcySsQOGPQklFyD1o9p9Oep05rz6M/A3sO5hIhz94fsM9ZjfvwHf\n/L9r4eTK5YZr0Nqov6kXZKqXFvqNg25+NcJJb7BxqH/fzF1qYKn+kvIf1bwvLWsn+bpJk4Q1WVvX\n/+uQ0Zaff3NbqoToZCS0CKvVXLvHvYsd7yRlsOWnHMyKOkfKyL7diR/Tl3EhvTi8fw8xUb619h8+\n6i71S8LL1fKb6qAfFqj/cSsKlBfV6MxpUDtxWu5nq/0WqqVtVm93/Z0b0bXoKOjs1Ms0Dm5qJ9Pu\n/cDR7epjN7Ufg0O32s/lHISPH4cRc9Xf1PuPu7HLK9D8L636LmH5j276+a3d31ptXX9n//yFaEck\ntAirVJrMeB39FzM9ezF9ZQXmqxcbp7pn8ZDXBfpM+APdu9g1fACzSe3sOeZ5WP+I2j8kOxl6DlJH\nXxRmQ8WvprTW26vDZV19oN/daitGRSmkvAeDJqkdMsf9FXyir3ba/HUH1Ho6oZ7dC5uf4ozHnfTJ\n/xYmvNW8ywMfP37jv6nLl1bbks9fiHZDQototgvF5SSdyCPpxAV2n8xnUKUNb9v8mQzn/+HTy4H8\nJewijxn+ASNWg50C+afU1pEiA36Zu6Fgw9WRM2fh8rnafT8yd6mhxHQFuvmrXxyuPuq8HtXzezh5\n1O5UWt28P/UDdftBDza/T8GWufDQv8k8Y6bPHY+17uUBIYQQTSKhRTTKbFZIzSli5/EL7DxxgVRD\nEQA9Xey4N8yTO4LCyf6lN3/bvYDF7oG4nzyO0aU3jh/PVueyqKEPWijwVMOH79BrgaT8MvzwBgyZ\nCYfWqn1UWqt539o+BfKbuhBCtAoJLaLejrTfpOXyeeo59Fot3528QH5JBRoNRPp2ZdHd/bkjqAcD\nezigyfiWvB9fpUvW1zhQgVPJESptu3L4chcCAqPo6dv/aguJGk52HUpnzNj/ql1AdUvJw++rX/SB\n/9W6l1ckdAghRLsgoUVYOtL+6d5gLhRfYdOhHI7/ovYjcXWwYUx/D8YG9WB0fw/cHG3AkAKH/gVH\nN4LxIl1sumL0GY1D/n4YGodNyns4j/kdm6oGEj+mb61zKdrMugXI5RUhhBBNIKFF0N3Jjv49u/DM\nfw4DoNNoeCDcixkj+hDh2xW9TqtOHZ/8D0j9UB3Wq7eHAbEQ9hAOOjscNj4BD6+1tFAM+mgWg6as\nBvpe99yAtHQIIYRoEgktndjRc0W8teMUfY6/i4MukCG9R3LgbAFz7+jLs/1y4fQ7kOuhBpWcFECj\nBonR/wPB94O9i3qg79+QlhIhhBA3nYSWTuhwdiFvfZvO9mMXcLbTc2fEKJ5L/yPz8nU8O2Ys9nte\nxbz3U7RmE2CGnqFw10sQOhlcvOoeUFpKhBBCtAIJLZ3IgTMFvLkjne9O5uHqYMPzY32Y3q+S8xk5\nfFURygrtX9Hu/ytQySWzC8aQR/AZPRN6hrR16UIIIYSElg7hegv+3baAg0ePsTVpF2XnjzPe9hde\n9ryEt8mAdk8O7AFnoJ9Gi8bWFcoLIHwqx0NfIjWnmPieTeiTIoQQQrQCCS0dwKf5vbhn92PYxC7F\nPS8TtnyM+fCH5Ot70WXHqwxWjAwGsAHFxgmNQ39wH6WukePeH9z7o7mcAxvnWFapHRkxjZFj5PKO\nEEKIW4eElvaq/DLkHABDCqPyvufKlTJsNv03g66+XKg4c6LShfM2IfgOCGfI4GHY9gpC4+xZezZZ\nUFtlNs658WnohRBCiFYgoeVWcL3LO7cvBLMZLp4CQ7K6Lo8hBS6kAepCP93cB5Db935STmUQoznI\nO1X3s9LuMRbeNYApQ3ywt9Fd//wyT4oQQoh2QELLrcB7cO2WjRNbYWMcBN0Hax9UQ0p5obqtnSv4\nRMHACeqtdxSninV89PF64tjGsqqJzLb9licemYNNYJ+mnV9G/wghhGgHJLTcCvxHw4Mr4YOHwNYR\njBfV5w+vhx7BMPABdcVi36HQvZ+6KjFQZKzkje0nObn3S97SL+NZZSFdeofy3C/hvP3RLJj6bwke\nQgghOgwJLbeCsgL4/nWoKlN/+twOoxepLTD2rnU2N5kVNuw/yz++PkmBsYK/98rld4XPEjf9MSqy\nj2B75wzmfgB/+Hk3fhJahBBCdBASWtraxQxY9zBcOg22XWD4U5CyCrS6egPL3tMXWfJZGsfOX2ao\nvxt/vn8gu9ODmH11wcOkbNSFD6fNYJuhiPg2eEtCCCHEzSChpS1l/QAfTgNTFdg6wdQPGhy9Yygw\n8rcvj/PFz+fx7urA/00bzD2DeqHRaAjxqhtuRvZ1r7VqsxBCCNHeSWhpKz+tg0/nQzc/GHAP9Lur\n3tE7Ru+RrEjK4F+7TqPRwLN39SdudEDjI4KEEEKIDkbblI22bdvGgAEDCAwMZOnSpXVeLygoYOLE\niYSFhTF06FCOHDliea2wsJDJkycTFBREcHAwP/74Y8tV3x6ZzbB9CWx+EvqMhCe+gbtfqtNhVvEb\nxZYuU7jzH9/x5renGD+oF98+F8P8O/tJYBFCCNEpNdrSYjKZmDt3Lt988w0+Pj5ER0czYcIEBg4c\naNnmlVdeISIigk2bNnH8+HHmzp3Ljh07AFiwYAHjx4/n448/pqKiAqPRePPeza2uwgib4yFtCwyZ\nBbGvgc6GFd9lEHa1TwrAz4YinvvoJ07mljDI24W3Hokkys+tbWsXQggh2lijoSU5OZnAwEACAgIA\nmDp1Klu2bKkVWtLS0li8eDEAQUFBZGVlkZubi729Pbt27WL16tUA2NraYmtrexPeRjtQ/AusfwTO\nHYK7/woj5lpmpg3zcWXeukP89TeDSDqRx4cp2WiAuFEBLL4nCK1Wc/1jCyGEEJ1Ao6ElJycHX19f\ny2MfHx/27dtXa5vw8HA2btzIqFGjSE5O5syZMxgMBnQ6HR4eHjz++OMcPnyYIUOGsGzZMpycnOqc\nJyEhgYSEBAAMBgNJSUlWvrWbo6SkpNm1OZVkEvrzy9hUlpA26PdcrBgE331Xa5t7eys89cFBAGy0\nMDfCjginXHbtyr3p9bUmqc86Up91pD4h2rcW6Yi7ePFiFixYQEREBKGhoURGRqLT6aiqquLgwYO8\n9dZbDBs2jAULFrB06VJeeumlOseIi4sjLi4OgKioKGJiYlqitBaXlJTUvNpObINPXgA7F5j1NaGe\n4XU22XEsl492HMTZXs/l8iqejAlk4d0DWqe+Vib1WUfqs47UJ0T71mho8fb2Jjs72/LYYDDg7e1d\naxsXFxcSExMBUBQFf39/AgICMBqN+Pj4MGzYMAAmT55cb0feDklRYO//wVd/BM9weGQDuHjW2WxD\n8ln+sOln/Lo7cclYwfyxgazdd5bhfbvLkGUhhBCihkZHD0VHR5Oenk5mZiYVFRVs2LCBCRMm1Nqm\nsLCQiooKAFauXMno0aNxcXGhV69e+Pr6cuLECQB27NhRqy9Mh2WqhM+fga/+AMH3weNf1gksiqLw\nxvaTLN74M4O8XSkwVvB/0wbz7N0DWP5oJPPWHWJPRn4bvQEhhBDi1tNoS4ter2f58uWMGzcOk8nE\n7NmzCQkJYcWKFQDEx8dz7NgxZs6cqU50FhLCqlWrLPu/9dZbTJs2jYqKCgICAiwtMh1KzVWaywrV\nieFO71Sn45+yxrJWULUqk5kXNh9hw/5sJg/xwd/dicjeXS0tKyP7urP80UhSDUXS2iKEEEJc1aQ+\nLbGxscTGxtZ6Lj7+2gTxI0aM4OTJk/XuGxERQUpKihUltgPVqzSPWwq7X4OL6eqU/DHP1wksxooq\nnl53iB3HLzDvjkCeu7s/Gk3d0UEyo60QQghRm8yI2xL8R6tDmDfFgc5WDSzVU/LXcLHkCr/9dwqp\nhkJe+s0gZgzv00YFCyGEEO2PhBZrmSphxxLY8xY49YDSC3DbgjqB5exFIzMTkzlXWMY704cwLqRX\nGxUshBBCtE9NmsZfNKDwLLw3Xg0s/e8BxQSjf6eu0py5y7LZkZwiJr2zh0ulFXzwxDAJLEIIIcQN\nkJaWG3X8C3X9ILP5WlCpXpW5xirNuyqDeXLtAbo62rIhbhiBPZzbunIhhBCiXZKWlmbSmCth62LY\n8Ki6QnP8LrB1uhZYwLJK88/JO5m9ej+9uzux8amREliEEEIIK0hLS3MUZBF56PdQnA5D/1tdnVlv\nB7cvrLWZoii8c8aLvx+KYmRfN1bMGIKLvU0bFS2EEEJ0DBJamirtU9gyD8eqKnjofRh4bYK9mqs0\nm8wKSz47ypofzxDs6czqx4diq5cGLSGEEMJa8m3amKor8OX/wH9mQPe+pES9XiuwwLVVmpNOXGDu\nBwdZ8+MZ7G20vBA7UAKLEEII0UKkpeV6LmbAx4/D+cMwfC7814uUf7+nzmYj+7rz5tRIZiUmU2VW\ncLTVsXJmlEwOJ4QQQrQgCS0NObIRPp0PWh1MXQ9BsdfdPCOvhCqzAsATt/tLYBFCCCFamFy7AHXt\noOp5VSrL4LOFaguLgyvE7240sJwrLONvXx7DRqfh6aurNMtih0IIIUTLktAC19YO+mk9rPwvOJAI\nenuY8BZ07X3dXRVFYd66g5RXmXltcjjPySrNQgghxE0hoQXUeVV+s0KdLO7iKbBzhmkfQd+xje66\n7cgvHDxbyKPDfHkg0huovUqzEEIIIVqG9Gmpduk0oEBVOYycX2ftoPoUlVXy50+PMtDThb9MGFTr\nNVmlWQghhGhZ0tIC6qKHu18Drb7etYMa8vdtx8kvucKrD4ah18lHKYQQQtxM8k0LkPQ3KM2DmN/D\n2D+qU/J/NOu6wWV/1iU+2HeW2bf5E+rj2mqlCiGEEJ2VhBZFgUNrwdUXbn9Wfe7q2kHkHKx3lytV\nJn6/8We8uzrwzF39W69WIYQQohOTPi2ntkNJLvzmHdDWyHD+oxvs1/Kv705z6kIJibOicbKTj1AI\nIYRoDfKN+/0b4OINgyY3afPzJWaW/3iK+8O9uCOox00uTgghhBDVOvflIUMKnPkeRswFvW2jm5vN\nCquPXsHeRsv/u29gKxQohBBCiGqdO7T88AbYu8Lgx5q0+UcHsjlRYOaP9wbj4Wx3k4sTQgghRE1N\nCi3btm1jwIABBAYGsnTp0jqvFxQUMHHiRMLCwhg6dChHjhyp9brJZCIyMpL77ruvZapuCfnpcOxz\niJ6jTibXiAvF5fz1i2MM6KbloSjfVihQCCGEEDU1GlpMJhNz585l69atpKWlsX79etLS0mpt88or\nrxAREUFqaipr1qxhwYIFtV5ftmwZwcHBLVu5tfa8CXo7GBbfpM3/8lka5ZVmZoXYodFobnJxQggh\nhPi1RkNLcnIygYGBBAQEYGtry9SpU9myZUutbdLS0hg7Vp3yPigoiKysLHJzcwEwGAx88cUXPPHE\nEzeh/BtU/Asc3gAR06CLR6Obf3s8l89TzzNvbCCeXTr3FTUhhBCirTQ6eignJwdf32uXQ3x8fNi3\nb1+tbcLDw9m4cSOjRo0iOTmZM2fOYDAY6NmzJwsXLuTvf/87xcXF1z1PQkICCQkJgBp0kpKSbuDt\nNE1Axr/xNVWxTxdNeSPnKa9S+OP3ZXh10TBQY6CkpPSm1matkpISqc8KUp91pD7r3Or1CdHWWmTI\n8+LFi1mwYAERERGEhoYSGRmJTqfj888/p0ePHgwZMqTRf4hxcXHExcUBEBUVRUxMTEuUVld5Efw4\nA0IeYPg9jzS6+Uufp3GxPJNPnhzBkD5uJCUl3bzaWoDUZx2pzzpSn3Vu9fqEaGuNhhZvb2+ys7Mt\njw0GA97e3rW2cXFxITExEQBFUfD39ycgIIAPP/yQTz/9lC+//JLy8nIuX77M9OnTWbt2bQu/jWZI\nSYQrl+G2hY1ummooJPGHTKYP782QPm6tUJwQQgghGtJoB43o6GjS09PJzMykoqKCDRs2MGHChFrb\nFBYWUlFRAcDKlSsZPXo0Li4u/O1vf8NgMJCVlcWGDRsYO3Zs2waWqiuw9x0IiAGviOtvajKz+JOf\nce9ix+/GB7VKeUIIIYRoWKMtLXq9nuXLlzNu3DhMJhOzZ88mJCSEFStWABAfH8+xY8eYOXMmGo2G\nkJAQVq1addMLvyGHN0DJLzBxRaObvvdDJmnnL7Ni+mBc7G1aoTghhBBCXE+T+rTExsYSGxtb67n4\n+GtDhUeMGMHJkyeve4yYmJi2vVZrNqvDnD3D1ZaW68i+ZOSf35zkroE9GRfSq1XKE0IIIcT1dZ7x\nuye+gIun4LYFcJ15VhRF4Y+bj6DTaPjLAyEyJ4sQQghxi+gcoUVR1IURu/lB8AN1Xl7xXQZ7MvIB\n+PTwOXadzGNKlC9bfjrXyoUKIYQQoiGdI7Sc+QFyUmDk06Cre0UszMeVeesO8fXRX/jLZ2n09XDi\n05/OEebj2gbFCiGEEKI+LTJPyy3vh2Xg5KHOgFuPkX3dWf5oJLMS91NZZabSbGbF9CGM7OveyoUK\nIYQQoiEdv6Ul9yikfw3D/htsHBrcbKifG1UmMwowa4SfBBYhhBDiFtPxQ8sPy8DGCaJ+e93NPks9\nh1mBO4N7sHbfWUsfFyGEEELcGjp2aCk8Cz9/DENmgWPDM9ruycjnT5uPAvDkmL4sfzSSeesOSXAR\nQgghbiEdO7T8+LY6vHnEU9fdLNVQxMRIdWkCf3cnSx+XVENRa1QphBBCiCbouKHFeAkOroHQh8DV\n57qbxo/pi4KCi70eNydbQO2cGz+mb2tUKoQQQogm6LihJfldqDTCbfObtHlmfin+Hl1kMjkhhBDi\nFtUxQ0uFEZL/Bf3HQ4/gJu2SmVdKgLvTTS5MCCGEEDeqY4aWQ2vBeBFuW9ikzcsqTJwrKsdfQosQ\nQghxy+p4ocVUBT++Bb7DoM+IJu2SdbEUgAAPCS1CCCHErarjhZa0zepQ5ya2soDanwWQlhYhhBDi\nFtYxQsv3b0DmrmsLI7oPUCeU+/6NJu1eHVr8uktoEUIIIW5VHSO0eA+Gj2bBD29A7s9qB9xPHlef\nb4LTeaX0crHHya5zLMUkhBBCtEcd41vafzRMWQ3vTwTbLvDTWvWx/+gm7Z6ZXyKXhoQQQohbXMdo\naQHwDAf7rlBRoq4z1MTAAtVztEhoEUIIIW5lHSe0nD8MKDDqOUhZpfZxaYKC0goKjJUyR4sQQghx\ni+sYoSVzl9qnZcpquPP/qbcfzWpScDktI4eEEEKIdqFjhJacg7X7sFT3cck52OiuMtxZCCGEaB+a\nFFq2bdvGgAEDCAwMZOnSpXVeLygoYOLEiYSFhTF06FCOHDkCQHZ2NnfccQcDBw4kJCSEZcuWtWz1\n1W5fWLcPi/9o9flGZOaXoNNq8HVzvDm1CSGEEKJFNBpaTCYTc+fOZevWraSlpbF+/XrS0tJqbfPK\nK68QERFBamoqa9asYcGCBQDo9Xr+8Y9/kJaWxt69e3n77bfr7NvWMvNL6e3miI2uYzQ6CSGEEB1V\no9/UycnJBAYGEhAQgK2tLVOnTmXLli21tklLS2Ps2LEABAUFkZWVRW5uLp6engwerM6V4uzsTHBw\nMDk5OTfhbdy403mlcmlICCGEaAcanaclJycHX19fy2MfHx/27dtXa5vw8HA2btzIqFGjSE5O5syZ\nMxgMBnr27GnZJisri0OHDjFs2LB6z5OQkEBCQgIABoOBpKSkG3k/zWJWFE5fMNLHrqzJ5yspKWmV\n2m6U1Gcdqc86Up91bvX6hGhrLTK53OLFi1mwYAERERGEhoYSGRmJTqezvF5SUsKDDz7IG2+8gYuL\nS73HiIuLIy4uDoCoqChiYmJaorTrOldYRsVX3zJ6cBAxw/o0aZ+kpKRWqe1GSX3WkfqsI/VZ51av\nT4i21mho8fb2Jjs72/LYYDDg7e1daxsXFxcSExMBUBQFf39/AgICAKisrOTBBx9k2rRpTJo0qSVr\nt5qMHBJCCCHaj0b7tERHR5Oenk5mZiYVFRVs2LCBCRMm1NqmsLCQiooKAFauXMno0aNxcXFBURR+\n+9vfEhwczLPPPntz3oEVqudoCXDv0saVCCGEEKIxjba06PV6li9fzrhx4zCZTMyePZuQkBBWrFgB\nQHx8PMeOHWPmzJloNBpCQkJYtWoVAD/88APvv/8+oaGhREREAOpIo9jY2Jv4lpouM68UBxsdPV3s\n2roUIYQQQjSiSX1aYmNj6wSN+Ph4y/0RI0Zw8uTJOvvdfvvtKIpiZYk3T/VCiRqNpq1LEUIIIUQj\nOvXkJLJQohBCCNF+dNrQUlFlJrugTBZKFEIIIdqJThtazl4yYjIrMnJICCGEaCc6bWiR4c5CCCFE\n+9KJQ0sJIKFFCCGEaC86cWgpxc3Jlq6Otm1dihBCCCGaoNOGFlkoUQghhGhfOm1oycyX0CKEEEK0\nJ50ytJRcqeJC8RUCZI4WIYQQot3olKEly7LmkIQWIYQQor3olKHltGW4syyUKIQQQrQXnTK0ZOaV\notFAn+6ObV2KEEIIIZqoc4aW/BK8XB2wt9G1dSlCCCGEaKJOGlpKpROuEEII0c50utCiKIrM0SKE\nEEK0Q50utOSXVFB8pUpCixBCCNHOdLrQIgslCiGEEO1TJwwt6kKJATLcWQghhGhXOl1oOZ1fio1O\ng3c3h7YuRQghhBDN0OlCS2ZeKX26O6HTatq6FCGEEEI0Q5NCy7Zt2xgwYACBgYEsXbq0zusFBQVM\nnDiRsLAwhg4dypEjR5q8b2uThRKFEEKI9qnR0GIymZg7dy5bt24lLS2N9evXk5aWVmubV155hYiI\nCFJTU1mzZg0LFixo8r6tyWRWOHPRKHO0CCGEEO1Qo6ElOTmZwMBAAgICsLW1ZerUqWzZsqXWNmlp\naYwdOxaAoKAgsrKyyM3NbdK+relcYRkVJrMslCiEEEK0Q42GlpycHHx9fS2PfXx8yMnJqbVNeHg4\nGzduBNSQc+bMGQwGQ5P2bU2yUKIQQgjRfulb4iCLFy9mwYIFREREEBoaSmRkJDpd89b1SUhIICEh\nAQCDwUBSUlJLlFbLN2cqATh/8jBJZ26sI25JSclNqa2lSH3WkfqsI/VZ51avT4i21mho8fb2Jjs7\n2/LYYDDg7e1daxsXFxcSExMBdZp8f39/AgICKCsra3TfanFxccTFxQEQFRVFTExMs99MY3ZuOYKz\nXQ4T7o5Bo7mx0JKUlHRTamspUp91pD7rSH3WudXrE6KtNXp5KDo6mvT0dDIzM6moqGDDhg1MmDCh\n1jaFhYVUVFQAsHLlSkaPHo2Li0uT9m1Np/NL8fdwuuHAIoQQQoi202hLi16vZ/ny5YwbNw6TycTs\n2bMJCQlhxYoVAMTHx3Ps2DFmzpyJRqMhJCSEVatWXXfftnI6r5Qov25tdn4hhBBC3Lgm9WmJjY0l\nNja21nPx8fGW+yNGjODkyZNN3rctlFeaOFdUhr+7T1uXIoQQQogb0GlmxD1z0YiiyEKJQgghRHvV\naUKLLJQohBBCtG+dJrRUz9Hi5+7YxpUIIYQQ4kZ0mtCSmVeKh7MdzvY2bV2KEEIIIW5A5wktslCi\nEEII0a51qtDSVxZKFEIIIdqtThFaioyVXCytkJYWIYQQoh3rFKEl86IslCiEEEK0d50jtFwd7iwt\nLUIIIUT71TlCS14pWg30dpPhzkIIIUR71SlCy+n8UnzdHLHVd4q3K4QQQnRIneJb/HSeDHcWQggh\n2rsOH1oURZE5WoQQQogOoMOHltzLVyirNBEgoUUIIYRo1zp8aDltGTkkw52FEEKI9qzDh5bMqwsl\n+stsuEIIIUS71vFDS14pdnotni72bV2KEEIIIazQ8UPL1U64Wq2mrUsRQgghhBU6RWgJkEtDQggh\nRLvXoUNLpcnM2UtGGe4shBBCdAAdOrQYCsqoMisyckgIIYToAJoUWrZt28aAAQMIDAxk6dKldV4v\nKiri/vvvJzw8nJCQEBITEy2vvf7664SEhDBo0CAeeeQRysvLW676RshCiUIIIUTH0WhoMZlMzJ07\nl61bt5KWlsb69etJS0urtc3bb7/NwIEDOXz4MElJSTz33HNUVFSQk5PDm2++SUpKCkeOHMFkMrFh\nw4ab9mZ+7XSeOtxZJpYTQggh2r9GQ0tycjKBgYEEBARga2vL1KlT2bJlS61tNBoNxcXFKIpCSUkJ\nbm5u6PV6AKqqqigrK6Oqqgqj0YiXl9fNeSf1yMwvpaujDd2cbFvtnEIIIYS4ORoNLTk5Ofj6+loe\n+/j4kJOTU2ubefPmcezYMby8vAgNDWXZsmVotVq8vb1ZtGgRvXv3xtPTE1dXV+6+++6WfxcNkIUS\nhRBCiI5D3xIH+eqrr4iIiODbb78lIyODu+66i1GjRmEymdiyZQuZmZl07dqVKVOmsHbtWqZPn17n\nGAkJCSQkJABgMBhISkqyuq7jOUYGdte1yLGqlZSUtOjxWprUZx2pzzpSn3Vu9fqEaGuNhhZvb2+y\ns7Mtjw0GA97e3rW2SUxMZPHixWg0GgIDA/H39+f48eOcOXMGf39/PDw8AJg0aRJ79uypN7TExcUR\nFxcHQFRUFDExMda8L0qvVFGw7StGDAogJqafVceqKSkpyerabiapzzpSn3WkPuvc6vUJ0dYavTwU\nHR1Neno6mZmZVFRUsGHDBiZMmFBrm969e7Njxw4AcnNzOXHiBAEBAfTu3Zu9e/diNBpRFIUdO3YQ\nHBx8c97Jr2RdvLrmkAx3FkIIITqERlta9Ho9y5cvZ9y4cZhMJmbPnk1ISAgrVqwAID4+nj/96U/M\nmjWL0NBQFEXh1Vdfxd3dHXd3dyZPnszgwYPR6/VERkZaWlNuNstCidKnRQghhOgQmtSnJTY2ltjY\n2GRKrFsAAA4JSURBVFrPxcfHW+57eXnx9ddf17vvkiVLWLJkiRUl3pjMq8Od/dwdW/3cQgghhGh5\nHXZG3Mz8Ujxd7XG0bZG+xkIIIYRoYx02tJyWhRKFEEKIDqVDhhZFUTidVyL9WYQQQogOpEOGlgJj\nJZfLq2TkkBBCCNGBdMjQUr1Qoqw5JIQQQnQcHTK0VC+UKJeHhBBCiI6jQ4aWzPxS9FoNPt0c2roU\nIYQQQrSQDhlaTueV0ru7I3pdh3x7QgghRKfUIb/VM/NLpT+LEEII0cF0uNBiNitkXiyV/ixCCCFE\nB9PhQsu5ojIqqswy3FkIIYToYDpcaJGFEoUQQoiOqcOGFpnCXwghhOhYOlxoOZ1XipOtjh7Odm1d\nihBCCCFaUIcLLZn5pfh7OKHRaNq6FCGEEEK0oI4ZWqQTrhBCCNHhdKjQcqXKhKHAKJ1whRBCiA6o\nQ4WW7EtGzIoslCiEEEJ0RB0qtMhCiUIIIUTH1aFCS/VwZz8JLUIIIUSH0yFCy4rvMtiTkc/pvFLc\nu9ji6mDDnox8VnyX0dalCSGEEKKFNCm0bNu2jQEDBhAYGMjSpUvrvF5UVMT9999PeHg4ISEhJCYm\nWl4rLCxk8uTJBAUFERwczI8//thy1f//9u42KKqyDwP4tYDUkwhFyIssScsywQLLxvs0TkyhJlQU\nqCOFIwXTQGFWOvqpFy0asr7gKGPDNBmWCumUOBHmxEgSZhgBzkAyiFCARqhBoD3ysv/nQ7WP++wi\n+izu2dXr94lz7vvsuc7tPYe/5xz2/E2v9sLq3c1o7R3CvT6zcbTrHFbvboZe7TXj+yIiIiJlTFu0\nTE5OorCwEDU1NWhvb8eePXvQ3t5u1qe0tBQ6nQ6tra2oq6vDunXrMDY2BgB46aWXsGTJEpw8eRKt\nra0IDw+f8YN4IMQH256+Hx0DI7h4eQKrdzdj29P344EQnxnfFxERESlj2qKlsbERWq0WGo0G7u7u\nyMrKQlVVlVkflUqFkZERiAhGR0fh7e0NNzc3DA8P48iRI8jLywMAuLu7484777whB5IQ7I37/Oeg\n/ewIVibew4KFiIjoJuM2XYf+/n4EBQWZltVqNb7//nuzPqtXr0Z6ejrmzZuHkZERVFZWwsXFBd3d\n3Zg7dy6effZZtLa2IjY2Flu2bMHs2ZYPypaVlaGsrAwA0NfXh7q6uus6kJ/OT6L//L+RHjILO+pP\n4V8jfQi/2/W6PuNajI6OXnc2e2I+2zCfbZjPNo6ej0hxMo29e/dKXl6eaXnnzp1SWFho0efll18W\no9EonZ2dEhwcLMPDw3L8+HFxdXWVY8eOiYjImjVr5NVXX51ulxIbGzttnys1nBqU+988JA2nBq0u\nz6TDhw/P+GfOJOazDfPZhvlsY0u+6z1vEjmjaW8PBQYGore317Tc19eHwMBAsz47duxAZmYmVCoV\ntFot7r33Xpw8eRJqtRpqtRqJiYkAgGXLluHHH3+c4bILONE3bPYMyz/PuJzoG57xfREREZEypi1a\n4uPj0dnZie7uboyNjaGiogLp6elmfe655x7U1tYCAAYGBtDR0QGNRgN/f38EBQWho6MDAFBbWwud\nTjfjB1GQHGLxDMsDIT4oSA6Z8X0RERGRMqZ9psXNzQ3btm3DI488gsnJSeTm5iIiIgLvv/8+AKCg\noACvvfYannnmGURFRUFEsHnzZvj4/FVEbN26FdnZ2RgbG4NGozH7c2giIiKiazVt0QIAaWlpSEtL\nM1tXUFBg+nnevHk4dOiQ1W0NBgN++OEHGyISERER3STfiEtEREQ3PxYtRERE5BRYtBAREZFTUImI\nKB3if/n4+CA4OFjpGFYNDg5i7ty5SseYEvPZhvlsw3y2sSVfT08Pzp07N8OJiByLQxYtjiwuLs6h\nHyxmPtswn22YzzaOno9Iabw9RERERE6BRQsRERE5BdeNGzduVDqEs4mNjVU6wlUxn22YzzbMZxtH\nz0ekJD7TQkRERE6Bt4eIiIjIKbBosaK3txcPPfQQdDodIiIisGXLFos+dXV18PLygsFggMFgwJtv\nvmnXjMHBwYiKioLBYEBcXJxFu4hgzZo10Gq10Ov1N+Tt2lPp6OgwjYvBYICnpydKSkrM+th7/HJz\nc+Hr64vIyEjTugsXLmDRokUIDQ3FokWL8Pvvv1vd9uDBg7jvvvug1Wrxzjvv2C3f+vXrERYWBr1e\nj4yMDAwNDVnddrq5cKPybdy4EYGBgaZ/wy+//NLqtkqN34oVK0zZgoODYTAYrG5rj/Gb6pziSHOQ\nyCkIWThz5ow0NTWJiMgff/whoaGh0tbWZtbn8OHD8uijjyoRT0RE5s+fL4ODg1O2V1dXy5IlS8Ro\nNMp3330nCQkJdkz3XxMTE+Ln5yc9PT1m6+09ft988400NTVJRESEad369euluLhYRESKi4tlw4YN\nFttNTEyIRqORrq4uuXz5suj1eou5cKPyffXVVzI+Pi4iIhs2bLCaT2T6uXCj8r3xxhvy3nvvXXU7\nJcfvSmvXrpVNmzZZbbPH+E11TnGkOUjkDHilxYqAgADExMQAAObMmYPw8HD09/crnOr6VFVVYdWq\nVVCpVEhKSsLQ0BDOnj1r9xy1tbUICQnB/Pnz7b7vKz344IPw9vY2W1dVVYWcnBwAQE5ODvbv32+x\nXWNjI7RaLTQaDdzd3ZGVlYWqqiq75Fu8eDHc3P56p2lSUhL6+vpmfL/Xylq+a6Hk+P1DRPDpp5/i\nqaeemvH9XqupzimONAeJnAGLlmn09PSgubkZiYmJFm1Hjx6FXq9Hamoq2tra7JpLpVJh4cKFiI2N\nRVlZmUV7f38/goKCTMtqtVqRwquiomLKXxZKjh8ADAwMICAgAADg7++PgYEBiz6OMo4ffvghUlNT\nrbZNNxdupK1bt0Kv1yM3N9fqrQ1HGL/6+nr4+fkhNDTUaru9x+/Kc4ozzUEiR8Ci5SpGR0exdOlS\nlJSUwNPT06wtJiYGv/zyC06cOIEXX3wRTz75pF2zffvtt2hpaUFNTQ1KS0tx5MgRu+7/WoyNjeHA\ngQNYvny5RZvS4/e/VCoVVCqVohmm8vbbb8PNzQ3Z2dlW25WaC88//zxOnz6NlpYWBAQEYN26dXbZ\n7/Xas2fPVa+y2HP8rnZOceQ5SOQoWLRMYXx8HEuXLkV2djYyMzMt2j09PeHh4QEASEtLw/j4uF3f\n+xEYGAgA8PX1RUZGBhobGy3ae3t7Tct9fX2mbeylpqYGMTEx8PPzs2hTevwAwM/Pz3TL7OzZs/D1\n9bXoo/Q4fvTRR/jiiy+wa9euKX+hTTcXbhQ/Pz+4urrCxcUFzz33nNX9Kj1+ExMT+Oyzz7BixYop\n+9hr/KydU5xhDhI5EhYtVogI8vLyEB4ejrVr11rt8+uvv0L+/oqbxsZGGI1G3H333XbJd/HiRYyM\njJh+PnTokNlfTQBAeno6du7cCRHBsWPH4OXlZboMbS9X+x+ukuP3j/T0dJSXlwMAysvL8cQTT1j0\niY+PR2dnJ7q7uzE2NoaKigqkp6fbJd/Bgwfx7rvv4sCBA7jjjjus9rmWuXCjXPmM1Oeff251v0qO\nHwB8/fXXCAsLg1qtttpur/Gb6pzi6HOQyOEo+RSwo6qvrxcAEhUVJdHR0RIdHS3V1dWyfft22b59\nu4iIbN26VXQ6nej1eklMTJSGhga75evq6hK9Xi96vV50Op0UFRWJiJjlMxqN8sILL4hGo5HIyEg5\nfvy43fKJiIyOjoq3t7cMDQ2Z1ik5fllZWeLv7y9ubm4SGBgoH3zwgZw7d04efvhh0Wq1kpKSIufP\nnxcRkf7+fklNTTVtW11dLaGhoaLRaExjbY98ISEholarTXMwPz/fIt9Uc8Ee+VauXCmRkZESFRUl\njz/+uJw5c8Yin4hy4ycikpOTY5pz/1Bi/KY6pzjSHCRyBvxGXCIiInIKvD1EREREToFFCxERETkF\nFi1ERETkFFi0EBERkVNg0UJEREROgUUL0Qyqq6vDY489pnQMIqKbEosWIiIicgosWuiW9MknnyAh\nIQEGgwH5+fmYnJyEh4cHXnnlFURERCAlJQWDg4MAgJaWFiQlJUGv1yMjI8P0YsBTp05h4cKFiI6O\nRkxMDLq6ugD89X6ZZcuWISwsDNnZ2eBXIRERzQwWLXTL+emnn1BZWYmGhga0tLTA1dUVu3btwsWL\nFxEXF4e2tjYkJydj06ZNAIBVq1Zh8+bNOHHiBKKiokzrs7OzUVhYiNbWVhw9etT0moTm5maUlJSg\nvb0dp0+fRkNDg2LHSkR0M3FTOgCRvdXW1qKpqQnx8fEAgD///BO+vr5wcXExvVhv5cqVyMzMxPDw\nMIaGhpCcnAwAyMnJwfLlyzEyMoL+/n5kZGQAAG6//XbT5yckJJjedWMwGNDT04MFCxbY8xCJiG5K\nLFroliMiyMnJQXFxsdn6t956y2x5qrcqT+e2224z/ezq6oqJiYn/63OIiMgcbw/RLSclJQX79u3D\nb7/9BgC4cOECfv75ZxiNRuzbtw8AsHv3bixYsABeXl646667UF9fDwD4+OOPkZycjDlz5kCtVmP/\n/v0AgMuXL+PSpUvKHBAR0S2CV1rolqPT6VBUVITFixfDaDRi1qxZKC0txezZs9HY2IiioiL4+vqi\nsrISAFBeXo6CggJcunQJGo0GO3bsAPBXAZOfn4/XX38ds2bNwt69e5U8LCKimx7f8kz0Nw8PD4yO\njiodg4iIpsDbQ0REROQUeKWFiIiInAKvtBAREZFTYNFCREREToFFCxERETkFFi1ERETkFFi0EBER\nkVNg0UJERERO4T/0iWkJ8D78mwAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "Image(filename='mnist_result/accuracy.png')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "もう少し学習を続ければ、まだ多少精度の向上が図れそうな雰囲気がありますね。\n",
    "\n",
    "ついでに、`dump_graph`という`Extension`が出力した計算グラフを、`Graphviz`を使って画像化して見てみましょう。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "%%bash\n",
    "dot -Tpng mnist_result/cg.dot -o mnist_result/cg.png"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "ここではJupyter notebook上でbashコマンドを使うCell magicを使用しています。2行目のコマンド自体は通常のシェルコマンドです。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAABMQAAATDCAYAAACZNelCAAAABmJLR0QA/wD/AP+gvaeTAAAgAElE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\nVf4/n14C/27u9lkAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "Image(filename='mnist_result/cg.png')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "上から下へ向かって、データやパラメータがどのような`Function`に渡されて計算が行われ、ロスを表す`Variable`が出力されたかが分かります。"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 8. 学習済みモデルで推論する"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAD8CAYAAAC4nHJkAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nH0iK8ANJEX4gKcIPJEX4gaQIP5DUXwFGhz+pWT5yuQAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f165810b940>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "label: 7\n",
      "predicted_label: 7\n"
     ]
    }
   ],
   "source": [
    "import numpy as np\n",
    "from chainer import serializers\n",
    "from chainer.cuda import to_gpu\n",
    "from chainer.cuda import to_cpu\n",
    "\n",
    "model = MLP()\n",
    "serializers.load_npz('mnist_result/model_epoch-10', model)\n",
    "if gpu_id >= 0:\n",
    "    model.to_gpu(gpu_id)\n",
    "\n",
    "x, t = test[0]\n",
    "plt.imshow(x.reshape(28, 28), cmap='gray')\n",
    "plt.show()\n",
    "print('label:', t)\n",
    "\n",
    "x = x[None, ...]\n",
    "x = to_gpu(x, gpu_id) if gpu_id >= 0 else x\n",
    "with chainer.using_config('train', False):\n",
    "    y = model(x)\n",
    "y = to_cpu(y.data)\n",
    "\n",
    "print('predicted_label:', y.argmax(axis=1)[0])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "無事正解できました。\n",
    "\n",
    "# 新しいネットワークを書いてみよう\n",
    "\n",
    "次に、MNISTデータセットではなくCIFAR10という32x32サイズの小さなカラー画像に10クラスのいずれかのラベルがついたデータセットを用いて、いろいろなモデルを自分で書いて試行錯誤する流れを体験してみます。\n",
    "\n",
    "| airplane | automobile | bird | cat | deer | dog | frog | horse | ship | truck |\n",
    "|:--------:|:----------:|:----:|:---:|:----:|:---:|:----:|:-----:|:----:|:-----:|\n",
    "| ![](https://www.cs.toronto.edu/~kriz/cifar-10-sample/airplane4.png) | ![](https://www.cs.toronto.edu/~kriz/cifar-10-sample/automobile4.png) | ![](https://www.cs.toronto.edu/~kriz/cifar-10-sample/bird4.png) | ![](https://www.cs.toronto.edu/~kriz/cifar-10-sample/cat4.png) | ![](https://www.cs.toronto.edu/~kriz/cifar-10-sample/deer4.png) | ![](https://www.cs.toronto.edu/~kriz/cifar-10-sample/dog4.png) | ![](https://www.cs.toronto.edu/~kriz/cifar-10-sample/frog4.png) | ![](https://www.cs.toronto.edu/~kriz/cifar-10-sample/horse4.png) | ![](https://www.cs.toronto.edu/~kriz/cifar-10-sample/ship4.png) | ![](https://www.cs.toronto.edu/~kriz/cifar-10-sample/truck4.png) |\n",
    "\n",
    "## 1. モデルの定義\n",
    "\n",
    "モデルは、`Chain`クラスを継承して定義します。ここでは、さきほど試した全結合層だけからなるネットワークではなく、畳込み層を持つネットワークを定義してみます。このモデルは3つの畳み込み層を持ち、2つの全結合層がそのあとに続いています。\n",
    "\n",
    "モデルの定義は主に2つのメソッドの定義によって行います。\n",
    "\n",
    "1. `__init__`コンストラクタでモデルを構成するレイヤーを定義する\n",
    "    - この際、親クラス（`Chain`）のコンストラクタに`super`を用いてキーワード引数としてモデルを構成する`Link`オブジェクトを渡すことで`Optimizer`から捕捉可能な最適化対象のパラメータを持つレイヤをモデルに追加することができます。\n",
    "2. データを受け取る`()`アクセサで呼ばれる`__call__`メソッドに、Forward計算を記述する"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import chainer\n",
    "import chainer.functions as F\n",
    "import chainer.links as L\n",
    "\n",
    "class MyModel(chainer.Chain):\n",
    "\n",
    "    def __init__(self, n_out):\n",
    "        super(MyModel, self).__init__()\n",
    "        with self.init_scope():\n",
    "            self.conv1 = L.Convolution2D(None, 6, 5, 1)\n",
    "            self.conv2 = L.Convolution2D(6, 16, 5, 1)\n",
    "            self.conv3 = L.Convolution2D(16, 120, 4, 1)\n",
    "            self.fc4 = L.Linear(None, 84)\n",
    "            self.fc5 = L.Linear(84, n_out)\n",
    "\n",
    "    def __call__(self, x):\n",
    "        h = F.relu(self.conv1(x))\n",
    "        h = F.max_pooling_2d(h, 2, 2)\n",
    "        h = F.relu(self.conv2(h))\n",
    "        h = F.relu(self.conv3(h))\n",
    "        h = F.relu(self.fc4(h))\n",
    "        h = self.fc5(h)\n",
    "        return h"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 2. 学習\n",
    "\n",
    "ここで、あとから別のモデルも簡単に同じ設定で訓練できるよう、`train`関数を定義しておきます。これは、\n",
    "\n",
    "- モデルのオブジェクト\n",
    "- バッチサイズ\n",
    "- 使用するGPU ID\n",
    "- 学習を終了するエポック数\n",
    "- データセットオブジェクト\n",
    "\n",
    "を渡すと、内部で`Trainer`を用いて渡されたデータセットを使ってモデルを訓練し、学習が終了した状態のモデルを返してくれる関数です。\n",
    "\n",
    "この`train`関数を用いて、上で定義した`MyModel`モデルを訓練してみます。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch       main/loss   main/accuracy  validation/main/loss  validation/main/accuracy  elapsed_time\n",
      "\u001b[J1           1.60486     0.419557       1.39821               0.502289                  5.4494        \n",
      "\u001b[J2           1.29682     0.538372       1.28784               0.540705                  11.1927       \n",
      "\u001b[J3           1.16459     0.586688       1.20128               0.57504                   17.0035       \n",
      "\u001b[J4           1.05643     0.629401       1.1574                0.58957                   22.7169       \n",
      "\u001b[J5           0.959835    0.661425       1.1393                0.597233                  28.4726       \n",
      "\u001b[J6           0.864292    0.696063       1.13649               0.620024                  34.1849       \n",
      "\u001b[J7           0.775904    0.726573       1.22777               0.600916                  40.0822       \n",
      "\u001b[J8           0.682157    0.760363       1.27081               0.601513                  45.9055       \n",
      "\u001b[J9           0.601456    0.786485       1.30436               0.602209                  51.7569       \n",
      "\u001b[J10          0.510369    0.818122       1.4064                0.596437                  57.3028       \n",
      "\u001b[J11          0.434679    0.846311       1.55172               0.587381                  62.7184       \n",
      "\u001b[J12          0.367371    0.869218       1.73642               0.599721                  68.3133       \n",
      "\u001b[J13          0.304524    0.892243       1.88752               0.572054                  74.2235       \n",
      "\u001b[J14          0.251589    0.910972       2.08466               0.574841                  80.162        \n",
      "\u001b[J15          0.216191    0.922935       2.22971               0.577528                  86.0273       \n",
      "\u001b[J16          0.203178    0.927537       2.40311               0.571855                  91.4411       \n",
      "\u001b[J17          0.181293    0.936441       2.45361               0.576931                  97.2668       \n",
      "\u001b[J18          0.156044    0.945963       2.64309               0.575239                  103.158       \n",
      "\u001b[J19          0.135993    0.953105       2.84978               0.573746                  108.645       \n",
      "\u001b[J20          0.142094    0.950684       2.98596               0.575438                  114.079       \n"
     ]
    }
   ],
   "source": [
    "from chainer.datasets import cifar\n",
    "from chainer import iterators\n",
    "from chainer import optimizers\n",
    "from chainer import training\n",
    "from chainer.training import extensions\n",
    "\n",
    "def train(model_object, batchsize=64, gpu_id=0,\n",
    "          max_epoch=20, train_dataset=None, test_dataset=None):\n",
    "\n",
    "    # 1. Dataset\n",
    "    if train_dataset is None and test_dataset is None:\n",
    "        train, test = cifar.get_cifar10()\n",
    "    else:\n",
    "        train, test = train_dataset, test_dataset\n",
    "\n",
    "    # 2. Iterator\n",
    "    train_iter = iterators.SerialIterator(train, batchsize)\n",
    "    test_iter = iterators.SerialIterator(test, batchsize, False, False)\n",
    "\n",
    "    # 3. Model\n",
    "    model = L.Classifier(model_object)\n",
    "    if gpu_id >= 0:\n",
    "        model.to_gpu(gpu_id)\n",
    "\n",
    "    # 4. Optimizer\n",
    "    optimizer = optimizers.Adam()\n",
    "    optimizer.setup(model)\n",
    "\n",
    "    # 5. Updater\n",
    "    updater = training.StandardUpdater(\n",
    "        train_iter, optimizer, device=gpu_id)\n",
    "\n",
    "    # 6. Trainer\n",
    "    trainer = training.Trainer(\n",
    "        updater, (max_epoch, 'epoch'),\n",
    "        out='{}_cifar10_result'.format(model_object.__class__.__name__))\n",
    "\n",
    "    trainer.extend(extensions.LogReport())\n",
    "    trainer.extend(extensions.Evaluator(test_iter, model, device=gpu_id))\n",
    "    trainer.extend(extensions.PrintReport(['epoch', 'main/loss', 'main/accuracy', 'validation/main/loss', 'validation/main/accuracy', 'elapsed_time']))\n",
    "    trainer.extend(extensions.PlotReport(['main/loss', 'validation/main/loss'], x_key='epoch', file_name='loss.png'))\n",
    "    trainer.extend(extensions.PlotReport(['main/accuracy', 'validation/main/accuracy'], x_key='epoch', file_name='accuracy.png'))\n",
    "    trainer.run()\n",
    "    del trainer\n",
    "\n",
    "    return model\n",
    "\n",
    "model = train(MyModel(10), gpu_id=0)  # CPUで実行する場合は、`gpu_id=-1`を指定して下さい。"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "学習が20エポックまで終わりました。ロスと精度のプロットを見てみましょう。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAg0AAAEKCAYAAACR9WaxAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nJAViaJAkSYEYGiRJUiB/AtBEvGPPnmGZAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "Image(filename='MyModel_cifar10_result/loss.png')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAicAAAEKCAYAAADJivWoAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nH6Qi1j2zWzc8AAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "Image(filename='MyModel_cifar10_result/accuracy.png')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "学習データでの精度（`main/accuracy`)は95%付近まで到達していますが、検証データでの精度（`validation/main/accuracy`）は一度60%程度に到達したあとは、下がっていってしまっています。検証データでのロス（`validation/main/loss`）を見ると、5エポック後以降はどんどん大きくなっていっていることが分かります。学習データでは良い精度が出ているが、検証データでは精度が良くないという場合、**モデルが学習データにオーバーフィッティングしている**ことが考えられます。\n",
    "\n",
    "## 3. 学習済みモデルを使った予測\n",
    "\n",
    "テスト精度は60%弱でしたが、試しにこの学習済みモデルを使っていくつかのテスト画像を分類させてみましょう。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "predicted_label: airplane\n",
      "answer: airplane\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAD8CAYAAAC4nHJkAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f16580d81d0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "predicted_label: truck\n",
      "answer: truck\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAD8CAYAAAC4nHJkAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f1659780ac8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "predicted_label: bird\n",
      "answer: dog\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAD8CAYAAAC4nHJkAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f165910d470>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "predicted_label: horse\n",
      "answer: horse\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAD8CAYAAAC4nHJkAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nrkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f1650960160>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "predicted_label: truck\n",
      "answer: truck\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAD8CAYAAAC4nHJkAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f16502cb6a0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "cls_names = ['airplane', 'automobile', 'bird', 'cat', 'deer',\n",
    "             'dog', 'frog', 'horse', 'ship', 'truck']\n",
    "\n",
    "def predict(model, image_id):\n",
    "    _, test = cifar.get_cifar10()\n",
    "    x, t = test[image_id]\n",
    "    model.to_cpu()\n",
    "    with chainer.using_config('train', False):\n",
    "        y = model.predictor(x[None, ...]).data.argmax(axis=1)[0]\n",
    "    print('predicted_label:', cls_names[y])\n",
    "    print('answer:', cls_names[t])\n",
    "\n",
    "    plt.imshow(x.transpose(1, 2, 0))\n",
    "    plt.show()\n",
    "\n",
    "for i in range(10, 15):\n",
    "    predict(model, i)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "分類に失敗しているものがいくつも見られます。モデルの学習に使用したデータセット上ではほぼ百発百中で正解できるとしても、未知のデータ、すなわち検証（or テスト）データセットにある画像に対して高精度な予測ができなければ、意味がありません[^NN]。検証データでの精度は、モデルの**汎化性能**に関係していると言われます。\n",
    "\n",
    "どうすれば高い汎化性能を持つモデルを設計し、学習することができるでしょうか？\n",
    "\n",
    "## 4. もっと深いモデルを定義してみよう\n",
    "\n",
    "では、上のモデルよりもよりたくさんの層を持つモデルを定義してみましょう。ここでは、1層の畳み込みネットワークを`ConvBlock`、1層の全結合ネットワークを`LinearBlock`として定義し、これをたくさんシーケンシャルに積み重ねる方法で大きなネットワークを定義してみます。\n",
    "\n",
    "### 構成要素を定義する\n",
    "\n",
    "まず、今目指している大きなネットワークの構成要素となる`ConvBlock`と`LinearBlock`を定義してみましょう。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "class ConvBlock(chainer.Chain):\n",
    "\n",
    "    def __init__(self, n_ch, pool_drop=False):\n",
    "        w = chainer.initializers.HeNormal()\n",
    "        super(ConvBlock, self).__init__()\n",
    "        with self.init_scope():\n",
    "            self.conv = L.Convolution2D(\n",
    "                None, n_ch, 3, 1, 1, nobias=True, initialW=w)\n",
    "            self.bn = L.BatchNormalization(n_ch)\n",
    "        self.pool_drop = pool_drop\n",
    "\n",
    "    def __call__(self, x):\n",
    "        h = F.relu(self.bn(self.conv(x)))\n",
    "        if self.pool_drop:\n",
    "            h = F.max_pooling_2d(h, 2, 2)\n",
    "            h = F.dropout(h, ratio=0.25)\n",
    "        return h\n",
    "\n",
    "class LinearBlock(chainer.Chain):\n",
    "\n",
    "    def __init__(self):\n",
    "        w = chainer.initializers.HeNormal()\n",
    "        super(LinearBlock, self).__init__()\n",
    "        with self.init_scope():\n",
    "            self.fc = L.Convolution2D(None, 1024, 1, 1, 0, initialW=w)\n",
    "\n",
    "    def __call__(self, x):\n",
    "        return F.dropout(F.relu(self.fc(x)), ratio=0.6)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "`ConvBlock`は`Chain`を継承したモデルとして定義されています。これは一つの畳み込み層とBatch Normalization層をパラメータありで持っているので、コンストラクタ内でこれらの登録を行っています。`__call__`メソッドでは、これらにデータを渡しつつ、活性化関数ReLUを適用して、さらに`pool_drop`がコンストラクタに`True`で渡されているときはMax PoolingとDropoutという関数を適用するような小さなネットワークになっています。\n",
    "\n",
    "Chainerでは、Pythonを使って書いたforward計算のコード自体がモデルの構造を表します。すなわち、実行時にデータがどのような層をくぐっていったか、ということがネットワークそのものを定義します。これによって、上記のような分岐などを含むネットワークも簡単に書け、柔軟かつシンプルで可読性の高いネットワーク定義が可能になります。これが**Define-by-Run**と呼ばれる特徴です。\n",
    "\n",
    "### 大きなネットワークの定義\n",
    "\n",
    "次に、これらの小さなネットワークを構成要素として積み重ねて、大きなネットワークを定義してみましょう。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "class DeepCNN(chainer.ChainList):\n",
    "\n",
    "    def __init__(self, n_output):\n",
    "        super(DeepCNN, self).__init__(\n",
    "            ConvBlock(64),\n",
    "            ConvBlock(64, True),  # 14\n",
    "            ConvBlock(128),\n",
    "            ConvBlock(128, True),  # 7\n",
    "            ConvBlock(256),\n",
    "            ConvBlock(256),\n",
    "            ConvBlock(256),\n",
    "            ConvBlock(512),\n",
    "            ConvBlock(512),\n",
    "            ConvBlock(512),\n",
    "            LinearBlock(),\n",
    "            LinearBlock(),\n",
    "            L.Linear(None, n_output)\n",
    "        )\n",
    "\n",
    "    def __call__(self, x):\n",
    "        for f in self:\n",
    "            x = f(x)\n",
    "        return x"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "ここで利用しているのが、`ChainList`というクラスです。このクラスは`Chain`を継承したクラスで、いくつもの`Link`や`Chain`を順次呼び出していくようなネットワークを定義するときに便利です。`ChainList`を継承して定義されるモデルは、親クラスのコンストラクタを呼び出す際に**キーワード引数ではなく普通の引数として**`Link`もしくは`Chain`オブジェクトを渡すことができます。そしてこれらは、**self.children()**メソッドによって**登録した順番に**取り出すことができます。\n",
    "\n",
    "この特徴を使うと、forward計算の記述が簡単になります。**self.children()**が返す構成要素のリストから、for文で構成要素を順番に取り出していき、そもそもの入力である`x`に取り出してきた部分ネットワークの計算を適用して、この出力で`x`を置き換えるということを順番に行っていけば、一連の`Link`または`Chain`を、コンストラクタで親クラスに登録した順番と同じ順番で適用していくことができます。そのため、シーケンシャルな部分ネットワークの適用によって表される大きなネットワークを定義するのに重宝します。\n",
    "\n",
    "それでは、学習を回してみます。今回はパラメータ数も多いので、学習を停止するエポック数を100に設定します。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch       main/loss   main/accuracy  validation/main/loss  validation/main/accuracy  elapsed_time\n",
      "\u001b[J1           1.85671     0.334739       1.39725               0.500398                  33.2824       \n",
      "\u001b[J2           1.19771     0.570122       1.0174                0.644208                  67.1486       \n",
      "\u001b[J3           0.943586    0.665993       0.818173              0.710191                  101.896       \n",
      "\u001b[J4           0.779686    0.728133       0.727823              0.753085                  137.454       \n",
      "\u001b[J5           0.67542     0.765145       0.671378              0.772492                  172.292       \n",
      "\u001b[J6           0.590576    0.797715       0.645285              0.779956                  207.662       \n",
      "\u001b[J7           0.527403    0.819462       0.573353              0.80215                   243.242       \n",
      "\u001b[J8           0.469255    0.838748       0.539177              0.818571                  278.583       \n",
      "\u001b[J9           0.419556    0.855579       0.498955              0.837381                  314.265       \n",
      "\u001b[J10          0.375575    0.870459       0.489314              0.835191                  349.89        \n",
      "\u001b[J11          0.34155     0.881782       0.519986              0.834594                  385.505       \n",
      "\u001b[J12          0.302928    0.895747       0.453357              0.856091                  420.344       \n",
      "\u001b[J13          0.272618    0.90671        0.471108              0.855096                  455.919       \n",
      "\u001b[J14          0.239327    0.916993       0.454362              0.860868                  491.484       \n",
      "\u001b[J15          0.213469    0.926376       0.500123              0.853205                  526.377       \n",
      "\u001b[J16          0.19075     0.934039       0.466309              0.858479                  562.143       \n",
      "\u001b[J17          0.176645    0.939198       0.495698              0.854697                  597.421       \n",
      "\u001b[J18          0.15366     0.947503       0.517126              0.864849                  632.684       \n",
      "\u001b[J19          0.142905    0.949844       0.506409              0.860569                  667.743       \n",
      "\u001b[J20          0.129063    0.955086       0.568722              0.860768                  703.249       \n",
      "\u001b[J21          0.115521    0.960218       0.521684              0.871716                  738.775       \n",
      "\u001b[J22          0.113413    0.960647       0.553135              0.857185                  773.868       \n",
      "\u001b[J23          0.104312    0.964749       0.533448              0.867436                  809.094       \n",
      "\u001b[J24          0.0968263   0.96743        0.52035               0.866441                  844.55        \n",
      "\u001b[J25          0.0881035   0.969889       0.53965               0.869228                  879.708       \n",
      "\u001b[J26          0.0839653   0.971071       0.505926              0.873905                  914.909       \n",
      "\u001b[J27          0.0800913   0.973391       0.524273              0.868133                  950.285       \n",
      "\u001b[J28          0.0726977   0.975632       0.525479              0.869526                  985.364       \n",
      "\u001b[J29          0.0691628   0.976363       0.550552              0.865744                  1020.93       \n",
      "\u001b[J30          0.06975     0.976873       0.568002              0.868133                  1055.85       \n",
      "\u001b[J31          0.0665702   0.977433       0.567897              0.873408                  1091.46       \n",
      "\u001b[J32          0.059455    0.979553       0.616676              0.86455                   1126.7        \n",
      "\u001b[J33          0.0598512   0.97924        0.582714              0.868929                  1162.05       \n",
      "\u001b[J34          0.0566354   0.980594       0.577343              0.866441                  1196.82       \n",
      "\u001b[J35          0.0581012   0.980614       0.556814              0.874701                  1232.1        \n",
      "\u001b[J36          0.0550008   0.981754       0.572455              0.865844                  1267.49       \n",
      "\u001b[J37          0.0495708   0.983596       0.566249              0.871318                  1302.65       \n",
      "\u001b[J38          0.0516591   0.982854       0.517448              0.875697                  1337.7        \n",
      "\u001b[J39          0.0482775   0.983335       0.565578              0.872711                  1373.21       \n",
      "\u001b[J40          0.0482059   0.983775       0.548205              0.873507                  1408.43       \n",
      "\u001b[J41          0.0442855   0.985334       0.527771              0.878284                  1443.83       \n",
      "\u001b[J42          0.0455125   0.984675       0.561503              0.872213                  1479.19       \n",
      "\u001b[J43          0.0460029   0.985335       0.542866              0.8751                    1514.4        \n",
      "\u001b[J44          0.0413705   0.986196       0.640892              0.869825                  1549.57       \n",
      "\u001b[J45          0.0430443   0.986073       0.593853              0.874104                  1585.2        \n",
      "\u001b[J46          0.0399809   0.987036       0.608239              0.873706                  1620.39       \n",
      "\u001b[J47          0.038206    0.987256       0.631329              0.8749                    1655.83       \n",
      "\u001b[J48          0.0408302   0.986496       0.611522              0.869128                  1691.2        \n",
      "\u001b[J49          0.0357845   0.988251       0.578226              0.872611                  1726.5        \n",
      "\u001b[J50          0.0366694   0.987796       0.588825              0.871815                  1762.17       \n"
     ]
    }
   ],
   "source": [
    "model = train(DeepCNN(10), max_epoch=50)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "学習が終了しました。ロスと精度のグラフを見てみましょう。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAhMAAAEKCAYAAACoktfqAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nTBAREZFb/gf92fTlS8fBGwAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "execution_count": 33,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "Image(filename='DeepCNN_cifar10_result/loss.png')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAicAAAEKCAYAAADJivWoAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "Image(filename='DeepCNN_cifar10_result/accuracy.png')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "先程よりも大幅にテストデータに対する精度が向上したことが分かります。62%前後だった精度が、87%程度まで上がりました。しかし最新の研究成果では97%近くまで達成されています。さらに精度を上げるには、今回行ったようなモデルの改良ももちろんのこと、学習データを擬似的に増やす操作（Data augmentation）や、複数のモデルの出力を一つの出力に統合する操作（Ensemble）などなど、いろいろな工夫が考えられます。\n",
    "\n",
    "# データセットクラスを書いてみよう\n",
    "\n",
    "ここでは、Chainerにすでに用意されているCIFAR10のデータを取得する機能を使って、データセットクラスを自分で書いてみます。Chainerでは、データセットを表すクラスは以下の機能を持っていることが必要とされます。\n",
    "\n",
    "- データセット内のデータ数を返す`__len__`メソッド\n",
    "- 引数として渡される`i`に対応したデータもしくはデータとラベルの組を返す`get_example`メソッド\n",
    "\n",
    "その他のデータセットに必要な機能は、`chainer.dataset.DatasetMixin`クラスを継承することで用意できます。ここでは、`DatasetMixin`クラスを継承し、Data augmentation機能のついたデータセットクラスを作成してみましょう。\n",
    "\n",
    "## 1. CIFAR10データセットクラスを書く"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "from chainer import dataset\n",
    "from chainer.datasets import cifar\n",
    "from chainercv import transforms\n",
    "\n",
    "class CIFAR10Augmented(dataset.DatasetMixin):\n",
    "\n",
    "    def __init__(self, train=True):\n",
    "        train_data, test_data = cifar.get_cifar10()\n",
    "        self.data = train_data if train else test_data\n",
    "        self.size = (28, 28)\n",
    "\n",
    "    def __len__(self):\n",
    "        return len(self.data)\n",
    "\n",
    "    def get_example(self, i):\n",
    "        x, t = self.data[i]\n",
    "        \n",
    "        # 学習モードなら画像にランダムな加工を加える\n",
    "        if chainer.config.train:\n",
    "            \n",
    "            # ランダムクロップ\n",
    "            x = transforms.random_crop(x, self.size) \n",
    "            \n",
    "            # ランダム左右反転\n",
    "            x = transforms.random_flip(x, x_random=True)\n",
    "            \n",
    "            # 色味を変える\n",
    "            x = transforms.pca_lighting(x, 0.1)\n",
    "        return x, t"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "このクラスは、CIFAR10のデータのそれぞれに対し、\n",
    "\n",
    "- 32x32の大きさの中からランダムに28x28の領域をクロップ\n",
    "- 1/2の確率で左右を反転させる\n",
    "\n",
    "という加工を行っています。こういった操作を加えることで擬似的に学習データのバリエーションを増やすと、オーバーフィッティングを抑制することに役に立つということが知られています。これらの操作以外にも、画像の色味を変化させるような変換やランダムな回転、アフィン変換など、さまざまな加工によって学習データ数を擬似的に増やす方法が提案されています。\n",
    "\n",
    "自分でデータの取得部分も書く場合は、コンストラクタに画像フォルダのパスとファイル名に対応したラベルの書かれたテキストファイルへのパスなどを渡してプロパティとして保持しておき、`get_example`メソッド内でそれぞれの画像を読み込んで対応するラベルとともに返す、という風にすれば良いことが分かります。\n",
    "\n",
    "## 2. 作成したデータセットクラスを使って学習を行う\n",
    "\n",
    "それではさっそくこの`CIFAR10`クラスを使って学習を行ってみましょう。先程使ったのと同じ大きなネットワークを使うことで、Data augmentationの効果がどの程度あるのかを調べてみましょう。`train`関数も含め、データセットクラス以外は先程使用したコードとほぼ同じになっています。異なるところはエポック数と、保存先ディレクトリ名だけです。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {
    "collapsed": false,
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch       main/loss   main/accuracy  validation/main/loss  validation/main/accuracy  elapsed_time\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Exception in main training loop: \n",
      "Invalid operation is performed in: LinearFunction (Forward)\n",
      "\n",
      "Expect: prod(in_types[0].shape[1:]) == in_types[1].shape[1]\n",
      "Actual: 65536 != 50176\n",
      "Traceback (most recent call last):\n",
      "  File \"/home/shunta/.pyenv/versions/anaconda3-4.3.1/lib/python3.6/site-packages/chainer-3.0.0a1-py3.6.egg/chainer/training/trainer.py\", line 313, in run\n",
      "    entry.extension(self)\n",
      "  File \"/home/shunta/.pyenv/versions/anaconda3-4.3.1/lib/python3.6/site-packages/chainer-3.0.0a1-py3.6.egg/chainer/training/extensions/evaluator.py\", line 137, in __call__\n",
      "    result = self.evaluate()\n",
      "  File \"/home/shunta/.pyenv/versions/anaconda3-4.3.1/lib/python3.6/site-packages/chainer-3.0.0a1-py3.6.egg/chainer/training/extensions/evaluator.py\", line 177, in evaluate\n",
      "    eval_func(*in_arrays)\n",
      "  File \"/home/shunta/.pyenv/versions/anaconda3-4.3.1/lib/python3.6/site-packages/chainer-3.0.0a1-py3.6.egg/chainer/links/model/classifier.py\", line 96, in __call__\n",
      "    self.y = self.predictor(*args, **kwargs)\n",
      "  File \"<ipython-input-31-41bbef2bfa65>\", line 22, in __call__\n",
      "    x = f(x)\n",
      "  File \"/home/shunta/.pyenv/versions/anaconda3-4.3.1/lib/python3.6/site-packages/chainer-3.0.0a1-py3.6.egg/chainer/links/connection/linear.py\", line 127, in __call__\n",
      "    return linear.linear(x, self.W, self.b)\n",
      "  File \"/home/shunta/.pyenv/versions/anaconda3-4.3.1/lib/python3.6/site-packages/chainer-3.0.0a1-py3.6.egg/chainer/functions/connection/linear.py\", line 102, in linear\n",
      "    return LinearFunction()(x, W, b)\n",
      "  File \"/home/shunta/.pyenv/versions/anaconda3-4.3.1/lib/python3.6/site-packages/chainer-3.0.0a1-py3.6.egg/chainer/function.py\", line 211, in __call__\n",
      "    ret = node.apply(inputs)\n",
      "  File \"/home/shunta/.pyenv/versions/anaconda3-4.3.1/lib/python3.6/site-packages/chainer-3.0.0a1-py3.6.egg/chainer/function_node.py\", line 206, in apply\n",
      "    self._check_data_type_forward(in_data)\n",
      "  File \"/home/shunta/.pyenv/versions/anaconda3-4.3.1/lib/python3.6/site-packages/chainer-3.0.0a1-py3.6.egg/chainer/function_node.py\", line 273, in _check_data_type_forward\n",
      "    self.check_type_forward(in_type)\n",
      "  File \"/home/shunta/.pyenv/versions/anaconda3-4.3.1/lib/python3.6/site-packages/chainer-3.0.0a1-py3.6.egg/chainer/function.py\", line 111, in check_type_forward\n",
      "    self._function.check_type_forward(in_types)\n",
      "  File \"/home/shunta/.pyenv/versions/anaconda3-4.3.1/lib/python3.6/site-packages/chainer-3.0.0a1-py3.6.egg/chainer/functions/connection/linear.py\", line 23, in check_type_forward\n",
      "    type_check.prod(x_type.shape[1:]) == w_type.shape[1],\n",
      "  File \"/home/shunta/.pyenv/versions/anaconda3-4.3.1/lib/python3.6/site-packages/chainer-3.0.0a1-py3.6.egg/chainer/utils/type_check.py\", line 514, in expect\n",
      "    expr.expect()\n",
      "  File \"/home/shunta/.pyenv/versions/anaconda3-4.3.1/lib/python3.6/site-packages/chainer-3.0.0a1-py3.6.egg/chainer/utils/type_check.py\", line 472, in expect\n",
      "    '{0} {1} {2}'.format(left, self.inv, right))\n",
      "Will finalize trainer extensions and updater before reraising the exception.\n"
     ]
    },
    {
     "ename": "InvalidType",
     "evalue": "\nInvalid operation is performed in: LinearFunction (Forward)\n\nExpect: prod(in_types[0].shape[1:]) == in_types[1].shape[1]\nActual: 65536 != 50176",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mInvalidType\u001b[0m                               Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-36-b66d39a53713>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[1;32m      1\u001b[0m model = train(\n\u001b[1;32m      2\u001b[0m     \u001b[0mDeepCNN\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;36m10\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mmax_epoch\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;36m50\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtrain_dataset\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mCIFAR10Augmented\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 3\u001b[0;31m     test_dataset=CIFAR10Augmented(False))\n\u001b[0m",
      "\u001b[0;32m<ipython-input-26-2c356918cb1b>\u001b[0m in \u001b[0;36mtrain\u001b[0;34m(model_object, batchsize, gpu_id, max_epoch, train_dataset, test_dataset)\u001b[0m\n\u001b[1;32m     41\u001b[0m     \u001b[0mtrainer\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mextend\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mextensions\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mPlotReport\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m'main/loss'\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m'validation/main/loss'\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mx_key\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;34m'epoch'\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfile_name\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;34m'loss.png'\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     42\u001b[0m     \u001b[0mtrainer\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mextend\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mextensions\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mPlotReport\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m'main/accuracy'\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m'validation/main/accuracy'\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mx_key\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;34m'epoch'\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfile_name\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;34m'accuracy.png'\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 43\u001b[0;31m     \u001b[0mtrainer\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mrun\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     44\u001b[0m     \u001b[0;32mdel\u001b[0m \u001b[0mtrainer\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     45\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/home/shunta/.pyenv/versions/anaconda3-4.3.1/lib/python3.6/site-packages/chainer-3.0.0a1-py3.6.egg/chainer/training/trainer.py\u001b[0m in \u001b[0;36mrun\u001b[0;34m(self, show_loop_exception_msg)\u001b[0m\n\u001b[1;32m    322\u001b[0m                 print('Will finalize trainer extensions and updater before '\n\u001b[1;32m    323\u001b[0m                       'reraising the exception.', file=sys.stderr)\n\u001b[0;32m--> 324\u001b[0;31m             \u001b[0msix\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mreraise\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0msys\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mexc_info\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    325\u001b[0m         \u001b[0;32mfinally\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    326\u001b[0m             \u001b[0;32mfor\u001b[0m \u001b[0m_\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mentry\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mextensions\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/home/shunta/.pyenv/versions/anaconda3-4.3.1/lib/python3.6/site-packages/six.py\u001b[0m in \u001b[0;36mreraise\u001b[0;34m(tp, value, tb)\u001b[0m\n\u001b[1;32m    684\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0mvalue\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m__traceback__\u001b[0m \u001b[0;32mis\u001b[0m \u001b[0;32mnot\u001b[0m \u001b[0mtb\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    685\u001b[0m             \u001b[0;32mraise\u001b[0m \u001b[0mvalue\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mwith_traceback\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mtb\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 686\u001b[0;31m         \u001b[0;32mraise\u001b[0m \u001b[0mvalue\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    687\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    688\u001b[0m \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/home/shunta/.pyenv/versions/anaconda3-4.3.1/lib/python3.6/site-packages/chainer-3.0.0a1-py3.6.egg/chainer/training/trainer.py\u001b[0m in \u001b[0;36mrun\u001b[0;34m(self, show_loop_exception_msg)\u001b[0m\n\u001b[1;32m    311\u001b[0m                     \u001b[0;32mfor\u001b[0m \u001b[0mname\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mentry\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mextensions\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    312\u001b[0m                         \u001b[0;32mif\u001b[0m \u001b[0mentry\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtrigger\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 313\u001b[0;31m                             \u001b[0mentry\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mextension\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    314\u001b[0m         \u001b[0;32mexcept\u001b[0m \u001b[0mException\u001b[0m \u001b[0;32mas\u001b[0m \u001b[0me\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    315\u001b[0m             \u001b[0;32mif\u001b[0m \u001b[0mshow_loop_exception_msg\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/home/shunta/.pyenv/versions/anaconda3-4.3.1/lib/python3.6/site-packages/chainer-3.0.0a1-py3.6.egg/chainer/training/extensions/evaluator.py\u001b[0m in \u001b[0;36m__call__\u001b[0;34m(self, trainer)\u001b[0m\n\u001b[1;32m    135\u001b[0m         \u001b[0;32mwith\u001b[0m \u001b[0mreporter\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    136\u001b[0m             \u001b[0;32mwith\u001b[0m \u001b[0mconfiguration\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0musing_config\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m'train'\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;32mFalse\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 137\u001b[0;31m                 \u001b[0mresult\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mevaluate\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    138\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    139\u001b[0m         \u001b[0mreporter_module\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mreport\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mresult\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/home/shunta/.pyenv/versions/anaconda3-4.3.1/lib/python3.6/site-packages/chainer-3.0.0a1-py3.6.egg/chainer/training/extensions/evaluator.py\u001b[0m in \u001b[0;36mevaluate\u001b[0;34m(self)\u001b[0m\n\u001b[1;32m    175\u001b[0m                 \u001b[0;32mwith\u001b[0m \u001b[0mfunction\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mno_backprop_mode\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    176\u001b[0m                     \u001b[0;32mif\u001b[0m \u001b[0misinstance\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0min_arrays\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtuple\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 177\u001b[0;31m                         \u001b[0meval_func\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0min_arrays\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    178\u001b[0m                     \u001b[0;32melif\u001b[0m \u001b[0misinstance\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0min_arrays\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mdict\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    179\u001b[0m                         \u001b[0meval_func\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m**\u001b[0m\u001b[0min_arrays\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/home/shunta/.pyenv/versions/anaconda3-4.3.1/lib/python3.6/site-packages/chainer-3.0.0a1-py3.6.egg/chainer/links/model/classifier.py\u001b[0m in \u001b[0;36m__call__\u001b[0;34m(self, *args, **kwargs)\u001b[0m\n\u001b[1;32m     94\u001b[0m         \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mloss\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     95\u001b[0m         \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0maccuracy\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 96\u001b[0;31m         \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0my\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mpredictor\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     97\u001b[0m         \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mloss\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlossfun\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0my\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mt\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     98\u001b[0m         \u001b[0mreporter\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mreport\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m{\u001b[0m\u001b[0;34m'loss'\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mloss\u001b[0m\u001b[0;34m}\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m<ipython-input-31-41bbef2bfa65>\u001b[0m in \u001b[0;36m__call__\u001b[0;34m(self, x)\u001b[0m\n\u001b[1;32m     20\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0m__call__\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mx\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     21\u001b[0m         \u001b[0;32mfor\u001b[0m \u001b[0mf\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 22\u001b[0;31m             \u001b[0mx\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mf\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mx\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     23\u001b[0m         \u001b[0;32mreturn\u001b[0m \u001b[0mx\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/home/shunta/.pyenv/versions/anaconda3-4.3.1/lib/python3.6/site-packages/chainer-3.0.0a1-py3.6.egg/chainer/links/connection/linear.py\u001b[0m in \u001b[0;36m__call__\u001b[0;34m(self, x)\u001b[0m\n\u001b[1;32m    125\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mW\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mdata\u001b[0m \u001b[0;32mis\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    126\u001b[0m             \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_initialize_params\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mx\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msize\u001b[0m \u001b[0;34m//\u001b[0m \u001b[0mx\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mshape\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m0\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 127\u001b[0;31m         \u001b[0;32mreturn\u001b[0m \u001b[0mlinear\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlinear\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mx\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mW\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mb\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[0;32m/home/shunta/.pyenv/versions/anaconda3-4.3.1/lib/python3.6/site-packages/chainer-3.0.0a1-py3.6.egg/chainer/functions/connection/linear.py\u001b[0m in \u001b[0;36mlinear\u001b[0;34m(x, W, b)\u001b[0m\n\u001b[1;32m    100\u001b[0m         \u001b[0;32mreturn\u001b[0m \u001b[0mLinearFunction\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mx\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mW\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    101\u001b[0m     \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 102\u001b[0;31m         \u001b[0;32mreturn\u001b[0m \u001b[0mLinearFunction\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mx\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mW\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mb\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[0;32m/home/shunta/.pyenv/versions/anaconda3-4.3.1/lib/python3.6/site-packages/chainer-3.0.0a1-py3.6.egg/chainer/function.py\u001b[0m in \u001b[0;36m__call__\u001b[0;34m(self, *inputs)\u001b[0m\n\u001b[1;32m    209\u001b[0m         \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_owned_node\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    210\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 211\u001b[0;31m         \u001b[0mret\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mnode\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mapply\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0minputs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    212\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    213\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0mlen\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mret\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;34m==\u001b[0m \u001b[0;36m1\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/home/shunta/.pyenv/versions/anaconda3-4.3.1/lib/python3.6/site-packages/chainer-3.0.0a1-py3.6.egg/chainer/function_node.py\u001b[0m in \u001b[0;36mapply\u001b[0;34m(self, inputs)\u001b[0m\n\u001b[1;32m    204\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    205\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0mconfiguration\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mconfig\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtype_check\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 206\u001b[0;31m             \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_check_data_type_forward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0min_data\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    207\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    208\u001b[0m         \u001b[0mhooks\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mchainer\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mget_function_hooks\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/home/shunta/.pyenv/versions/anaconda3-4.3.1/lib/python3.6/site-packages/chainer-3.0.0a1-py3.6.egg/chainer/function_node.py\u001b[0m in \u001b[0;36m_check_data_type_forward\u001b[0;34m(self, in_data)\u001b[0m\n\u001b[1;32m    271\u001b[0m         \u001b[0min_type\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mtype_check\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mget_types\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0min_data\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m'in_types'\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;32mFalse\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    272\u001b[0m         \u001b[0;32mwith\u001b[0m \u001b[0mtype_check\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mget_function_check_context\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 273\u001b[0;31m             \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcheck_type_forward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0min_type\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    274\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    275\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0mcheck_type_forward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0min_types\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/home/shunta/.pyenv/versions/anaconda3-4.3.1/lib/python3.6/site-packages/chainer-3.0.0a1-py3.6.egg/chainer/function.py\u001b[0m in \u001b[0;36mcheck_type_forward\u001b[0;34m(self, in_types)\u001b[0m\n\u001b[1;32m    109\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    110\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0mcheck_type_forward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0min_types\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 111\u001b[0;31m         \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_function\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcheck_type_forward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0min_types\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    112\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    113\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0mforward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0minputs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/home/shunta/.pyenv/versions/anaconda3-4.3.1/lib/python3.6/site-packages/chainer-3.0.0a1-py3.6.egg/chainer/functions/connection/linear.py\u001b[0m in \u001b[0;36mcheck_type_forward\u001b[0;34m(self, in_types)\u001b[0m\n\u001b[1;32m     21\u001b[0m             \u001b[0mx_type\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mndim\u001b[0m \u001b[0;34m>=\u001b[0m \u001b[0;36m2\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     22\u001b[0m             \u001b[0mw_type\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mndim\u001b[0m \u001b[0;34m==\u001b[0m \u001b[0;36m2\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 23\u001b[0;31m             \u001b[0mtype_check\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mprod\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mx_type\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mshape\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;34m==\u001b[0m \u001b[0mw_type\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mshape\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     24\u001b[0m         )\n\u001b[1;32m     25\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0mtype_check\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0meval\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mn_in\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;34m==\u001b[0m \u001b[0;36m3\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/home/shunta/.pyenv/versions/anaconda3-4.3.1/lib/python3.6/site-packages/chainer-3.0.0a1-py3.6.egg/chainer/utils/type_check.py\u001b[0m in \u001b[0;36mexpect\u001b[0;34m(*bool_exprs)\u001b[0m\n\u001b[1;32m    512\u001b[0m         \u001b[0;32mfor\u001b[0m \u001b[0mexpr\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mbool_exprs\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    513\u001b[0m             \u001b[0;32massert\u001b[0m \u001b[0misinstance\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mexpr\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mTestable\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 514\u001b[0;31m             \u001b[0mexpr\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mexpect\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    515\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    516\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/home/shunta/.pyenv/versions/anaconda3-4.3.1/lib/python3.6/site-packages/chainer-3.0.0a1-py3.6.egg/chainer/utils/type_check.py\u001b[0m in \u001b[0;36mexpect\u001b[0;34m(self)\u001b[0m\n\u001b[1;32m    470\u001b[0m             raise InvalidType(\n\u001b[1;32m    471\u001b[0m                 \u001b[0;34m'{0} {1} {2}'\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mformat\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlhs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mexp\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mrhs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 472\u001b[0;31m                 '{0} {1} {2}'.format(left, self.inv, right))\n\u001b[0m\u001b[1;32m    473\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    474\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mInvalidType\u001b[0m: \nInvalid operation is performed in: LinearFunction (Forward)\n\nExpect: prod(in_types[0].shape[1:]) == in_types[1].shape[1]\nActual: 65536 != 50176"
     ]
    }
   ],
   "source": [
    "model = train(\n",
    "    DeepCNN(10), max_epoch=50, train_dataset=CIFAR10Augmented(),\n",
    "    test_dataset=CIFAR10Augmented(False))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "先程のData augmentationなしの場合は87%程度で頭打ちになっていた精度が、学習データにaugmentationを施すことで89%以上まで向上させられることが分かりました。2%強の改善です。\n",
    "\n",
    "最後に、ロスと精度のグラフを見てみましょう。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "Image(filename='DeepCNN_cifar10_result/loss.png')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false,
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "Image(filename='DeepCNN_cifar10_result/accuracy.png')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# おわりに\n",
    "\n",
    "本記事では、[Chainer](http://chainer.org/)に関する\n",
    "\n",
    "- Trainerを使わない学習ループの書き方\n",
    "- Trainerの使い方\n",
    "- 自作モデルの書き方\n",
    "- 自作データセットクラスの書き方\n",
    "\n",
    "を簡単に紹介しました。今後もHands-on形式で行うかどうかは分かりませんが、以下のようなものの解説をどこかに書いていきたいと思います。\n",
    "\n",
    "- Trainerを構成するUpdaterやIteratorを自作する方法\n",
    "- `chainer.links.models.vision`モジュール以下にある`VGG16Layers`や`ResNet50Layers`というPre-trainedモデルを特定のタスクに向けてFine-tuningする方法\n",
    "- Extensionの作り方\n",
    "\n",
    "また、Chainerの開発にコミットしてくれる方を歓迎します！Chainerはオープンソースソフトウェアですので、皆さんが自身で欲しい機能などを提案し、Pull requestを送ることで進化していきます。興味のある方は、こちらの[Contoribution Guide](http://docs.chainer.org/en/latest/contribution.html)をお読みになった後、ぜひIssueを立てたりPRを送ったりしてみてください。お待ちしております。\n",
    "\n",
    "pfent/chainer\n",
    "[https://github.com/pfnet/chainer](https://github.com/pfnet/chainer)\n",
    "\n",
    "### 脚注\n",
    "\n",
    "[^TrainingデータとValidationデータ]: 本記事では、Chainerの使い方の説明に主眼を置いているため、ValidationデータセットとTestデータセットを明確に区別していません。しかし実際にはこれらは区別されるべきです。普通、Trainingデータの一部をTrainingデータセットから取り除き、それらの取り除かれたデータでValidationデータセットを構成しておきます。その後、Trainingデータで訓練したモデルをまずValidationデータで評価し、Validationデータでの性能を向上させるようにモデルを改良していくというのが一般的な手順です。Testデータは全ての取り組みが終了したあとに、最終的なそのモデルの性能を（例えば他のモデルなどと比較する目的で）評価するためにだけ用いられます。偏ったデータを使ってモデル改良を行ってしまいオーバーフィッティングなどに陥ることを避けるなどの目的で、Training/Validationデータの構成を複数用意しておく場合もあります。\n",
    "[^NN]: 学習データに対する予測精度は、もし学習データから抜き出されたあるデータをクエリとし、それが含まれている学習データセットから検索して発見することが必ずできるならば、そのデータについているラベルを答えることで、100%になってしまいます。"
   ]
  }
 ],
 "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": "3.6.0"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}