pytorch_nueral_net.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "#データ準備\n",
    "import pandas as pd\n",
    "from sklearn import datasets\n",
    "\n",
    "iris = datasets.load_iris()\n",
    "df = pd.DataFrame(iris.data, columns=iris.feature_names)\n",
    "df['target'] = iris.target_names[iris.target]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "import torch\n",
    "import numpy as np\n",
    "\n",
    "#ラベルを数値化\n",
    "y = np.array(df['target'].astype('category').cat.codes).astype(float)\n",
    "X = np.array(df.iloc[:, :4])\n",
    "\n",
    "#学習データと検証データを分割\n",
    "from sklearn.model_selection import train_test_split\n",
    "train_X, val_X, train_y, val_y = train_test_split(\n",
    "    X, y, test_size = 0.2, random_state=71)\n",
    "\n",
    "\n",
    "# tensor型に変換\n",
    "train_X = torch.Tensor(train_X)\n",
    "val_X = torch.Tensor(val_X)\n",
    "train_y = torch.LongTensor(train_y)\n",
    "val_y = torch.LongTensor(val_y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Net(\n",
      "  (fc1): Linear(in_features=4, out_features=100, bias=True)\n",
      "  (fc2): Linear(in_features=100, out_features=50, bias=True)\n",
      "  (fc3): Linear(in_features=50, out_features=3, bias=True)\n",
      ")\n"
     ]
    }
   ],
   "source": [
    "#ネットワーク定義\n",
    "import torch.nn as nn\n",
    "import torch.nn.functional as F\n",
    "from torch.autograd import Variable\n",
    "\n",
    "torch.manual_seed(71) #seed固定、ネットワーク定義前にする必要ありそう\n",
    "\n",
    "class Net(nn.Module):\n",
    "\n",
    "    def __init__(self):\n",
    "        super(Net, self).__init__()\n",
    "        self.fc1 = nn.Linear(4, 100)\n",
    "        self.fc2 = nn.Linear(100, 50)\n",
    "        self.fc3 = nn.Linear(50, 3)\n",
    "\n",
    "    def forward(self, x):\n",
    "        x = F.relu(self.fc1(x))\n",
    "        x = F.relu(self.fc2(x))\n",
    "        x = self.fc3(x)\n",
    "        return F.log_softmax(x, dim = 1)\n",
    "\n",
    "model = Net()\n",
    "print(model)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train Step: 0\tLoss: 1.152\tAccuracy: 0.317\n",
      "Train Step: 10\tLoss: 0.954\tAccuracy: 0.683\n",
      "Train Step: 20\tLoss: 0.835\tAccuracy: 0.683\n",
      "Train Step: 30\tLoss: 0.711\tAccuracy: 0.683\n",
      "Train Step: 40\tLoss: 0.616\tAccuracy: 0.692\n",
      "Train Step: 50\tLoss: 0.547\tAccuracy: 0.717\n",
      "Train Step: 60\tLoss: 0.496\tAccuracy: 0.783\n",
      "Train Step: 70\tLoss: 0.457\tAccuracy: 0.867\n",
      "Train Step: 80\tLoss: 0.425\tAccuracy: 0.917\n",
      "Train Step: 90\tLoss: 0.399\tAccuracy: 0.925\n",
      "Train Step: 100\tLoss: 0.377\tAccuracy: 0.942\n"
     ]
    }
   ],
   "source": [
    "#学習\n",
    "import torch.optim as optim\n",
    "\n",
    "optimizer = optim.SGD(model.parameters(), lr=0.02)\n",
    "train_loss = []\n",
    "train_accu = []\n",
    "i = 0\n",
    "\n",
    "model.train() #学習モード\n",
    "for epoch in range(100):\n",
    "    data, target = Variable(train_X), Variable(train_y)#微分可能な型\n",
    "    optimizer.zero_grad() #勾配初期化\n",
    "    output = model(data) #データを流す\n",
    "        \n",
    "    loss = F.nll_loss(output, target) #loss計算\n",
    "    loss.backward()    #バックプロパゲーション\n",
    "    train_loss.append(loss.data.item())\n",
    "    optimizer.step()   # 重み更新\n",
    "        \n",
    "    prediction = output.data.max(1)[1] #予測結果\n",
    "    accuracy = prediction.eq(target.data).sum().numpy() / len(train_X) #正解率\n",
    "    train_accu.append(accuracy)\n",
    "    \n",
    "    if i % 10 == 0:\n",
    "        print('Train Step: {}\\tLoss: {:.3f}\\tAccuracy: {:.3f}'.format(i, loss.data.item(), accuracy))\n",
    "    i += 1\n",
    "    \n",
    "print('Train Step: {}\\tLoss: {:.3f}\\tAccuracy: {:.3f}'.format(i, loss.data.item(), accuracy))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy: 0.867\n"
     ]
    }
   ],
   "source": [
    "#精度検証\n",
    "model.eval() #推論モード\n",
    "\n",
    "outputs = model(Variable(val_X))\n",
    "_, predicted = torch.max(outputs.data, 1)\n",
    "print('Accuracy: {:.3f}'.format(predicted.eq(val_y).sum().numpy() / len(predicted)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.5.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}