Neural Networks using PyTorch.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# MNIST Dataset Introduction\n",
    "A dataset of handwritten digits. The dataset contains 60,000 examples for training and 10,000 examples for testing. The digits have been size-normalized and centered in a fixed-size image (28x28 pixels) with values from 0 to 1.\n",
    "![MNIST](imgs/mnist.png)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import torch\n",
    "import torch.nn as nn\n",
    "import torch.utils as utils\n",
    "from torch.autograd import Variable\n",
    "import torchvision.datasets as datasets\n",
    "import torchvision.transforms as transforms"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "#The dataset is downloaded for you. this will transform it into PyTorch tensors\n",
    "mnist_train = datasets.MNIST(\"/var/local/mnist\", train=True, transform=transforms.ToTensor(), download=False)\n",
    "mnist_test = datasets.MNIST(\"/var/local/mnist\", train=False, transform=transforms.ToTensor(), download=False)\n",
    "def to_categorical(y):\n",
    "    # 1-hot encodes a tensor \n",
    "    return torch.from_numpy(np.eye(10, dtype='float32')[y])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Simple Neural Network\n",
    "![NN](imgs/nn.jpeg)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 92,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Set Data Loader(input pipeline)\n",
    "train_loader = torch.utils.data.DataLoader(dataset=mnist_train,batch_size=16,shuffle=True)\n",
    "test_loader = torch.utils.data.DataLoader(dataset=mnist_test,batch_size=16,shuffle=True)\n",
    "\n",
    "# create the model\n",
    "model = torch.nn.Sequential(\n",
    "    torch.nn.Linear(28*28, 32),\n",
    "    torch.nn.ReLU(),\n",
    "    torch.nn.Linear(32,16),\n",
    "    torch.nn.ReLU(),\n",
    "    torch.nn.Linear(16,10)\n",
    ")\n",
    "loss_func = nn.MultiLabelSoftMarginLoss()\n",
    "optimizer = torch.optim.Adam(model.parameters())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 90,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.6801751852035522\n",
      "0.11195375770330429\n",
      "0.1046844869852066\n",
      "0.20801544189453125\n",
      "0.06674367934465408\n",
      "0.099596306681633\n",
      "0.07238338887691498\n",
      "0.04316641017794609\n"
     ]
    }
   ],
   "source": [
    "# Train the model\n",
    "loss_arr=[]\n",
    "for i in range(1):\n",
    "    for j,[image,label] in enumerate(train_loader):\n",
    "        image = Variable(image.view(-1,1,28*28))\n",
    "        label = Variable(to_categorical(label).view(-1,1,10))\n",
    "        optimizer.zero_grad()\n",
    "        prediction = model.forward(image)\n",
    "        loss = loss_func(prediction,label)\n",
    "        loss.backward()\n",
    "        optimizer.step()\n",
    "        loss_arr.append(loss.data.tolist())\n",
    "        if j % 500 == 0:\n",
    "            print(loss.data.tolist())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 91,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f28814bccc0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy on the test data: 0.932\n"
     ]
    }
   ],
   "source": [
    "# Plot the training loss curve\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    "plt.plot(np.arange(len(loss_arr)), loss_arr, label='Loss')\n",
    "plt.legend()\n",
    "plt.show()\n",
    "\n",
    "# Calculate the accuracy\n",
    "model.eval()\n",
    "correct = 0\n",
    "total = 0\n",
    "\n",
    "for image,label in test_loader:\n",
    "    image = Variable(image.view(-1,1,28*28))\n",
    "    result = model(image)\n",
    "    pred = torch.argmax(result.data,2).view(-1) \n",
    "    total += label.size(0)\n",
    "    correct += (pred == label).sum().numpy()\n",
    "print(\"Accuracy on the test data: {}\".format(correct/total))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}