LSTM.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 336,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/opt/conda/lib/python2.7/site-packages/ipykernel_launcher.py:16: FutureWarning: convert_objects is deprecated.  To re-infer data dtypes for object columns, use DataFrame.infer_objects()\n",
      "For all other conversions use the data-type specific converters pd.to_datetime, pd.to_timedelta and pd.to_numeric.\n",
      "  app.launch_new_instance()\n"
     ]
    }
   ],
   "source": [
    "import pandas as pd \n",
    "import numpy as np\n",
    "import os\n",
    "from sklearn.model_selection import train_test_split\n",
    "import torch\n",
    "import torch.nn as nn\n",
    "import torchvision\n",
    "import torchvision.transforms as transforms\n",
    "from matplotlib import pyplot as plt\n",
    "import torch.utils.data as data_utils\n",
    "import itertools\n",
    "from sklearn.metrics import confusion_matrix\n",
    "\n",
    "path = \"/datasets/home/73/673/h6gupta/Challenge/challenge_dataset.csv\"\n",
    "data = pd.read_csv(path, header = None)\n",
    "data = data.convert_objects(convert_numeric=True)\n",
    "\n",
    "df1 = data.iloc[1:,2:16].values.reshape(205,89,14)\n",
    "df1 = np.moveaxis(df1, [1,2], [2,1])\n",
    "\n",
    "labels =[]\n",
    "for i in range(1,206):\n",
    "    labels.append(data.iloc[89*i,16])\n",
    "\n",
    "labels = np.asarray(labels).astype('int64')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 337,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train = (df1 - np.mean(df1, axis=0)[None,:,:]) / np.std(df1, axis=0)[None,:,:]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 338,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "torch.Size([174, 14, 89])\n",
      "torch.Size([174])\n",
      "<class 'torch.Tensor'>\n",
      "<class 'torch.Tensor'>\n"
     ]
    }
   ],
   "source": [
    "X_tr, X_vld, lab_tr, lab_vld = train_test_split(X_train, labels, stratify = labels, test_size = 0.15)\n",
    "y_tr = lab_tr\n",
    "y_vld = lab_vld\n",
    "\n",
    "img = torch.from_numpy(X_tr).float().cuda()\n",
    "lbl = torch.from_numpy(y_tr).long().cuda()\n",
    "\n",
    "\n",
    "batch_size = 40\n",
    "\n",
    "train = data_utils.TensorDataset(img,lbl)\n",
    "data_loader = data_utils.DataLoader(train, batch_size = batch_size, shuffle = True)\n",
    "print(img.size())\n",
    "print(lbl.size())\n",
    "print(type(img))\n",
    "print(type(lbl))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 342,
   "metadata": {},
   "outputs": [],
   "source": [
    "class LSTM(nn.Module):\n",
    "    def __init__(self, num_classes):\n",
    "        super(LSTM, self).__init__()\n",
    "        self.input_dim = 14\n",
    "        self.hidden_dim = 64\n",
    "        self.num_layers = 1\n",
    "\n",
    "        # Define the LSTM layer\n",
    "        self.lstm = nn.LSTM(self.input_dim, self.hidden_dim, self.num_layers)\n",
    "\n",
    "        # Define the output layer\n",
    "        self.linear = nn.Linear(self.hidden_dim, num_classes)\n",
    "        \n",
    "    def init_hidden(self, batch_size):\n",
    "        # This is what we'll initialise our hidden state as\n",
    "        return (torch.randn(self.num_layers, batch_size, self.hidden_dim).cuda(),\n",
    "                torch.randn(self.num_layers, batch_size, self.hidden_dim).cuda())\n",
    "        \n",
    "    def forward(self, X):\n",
    "        batch_size = X.shape[1]\n",
    "        output, self.hidden = self.lstm(X, self.hidden)\n",
    "        y = self.linear(output[-1].view(batch_size, -1))\n",
    "        #y = self.softmax(y)\n",
    "        return y        "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 343,
   "metadata": {},
   "outputs": [],
   "source": [
    "num_epochs = 1001\n",
    "model = LSTM(3).cuda()\n",
    "loss_function = nn.CrossEntropyLoss()\n",
    "optimizer = torch.optim.Adam(model.parameters(), lr=0.0001)\n",
    "\n",
    "def lr_decay(optimizer, epoch, init_lr=0.001, lr_decay_epoch=400):\n",
    "    lr = init_lr * (0.1**(epoch // lr_decay_epoch))\n",
    "    for param_group in optimizer.param_groups:\n",
    "        param_group['lr'] = lr\n",
    "    return optimizer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 344,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Testing Accuracy: 0.23\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "def accuracy(yinit_v1, lbl_v1):\n",
    "    return (torch.max(yinit_v1.data, 1)[1].long() == lbl_v1.data).sum().item()\n",
    "\n",
    "img_v1 = torch.from_numpy(X_vld).float().cuda()\n",
    "lbl_v1 = torch.from_numpy(y_vld).long().cuda()\n",
    "model.hidden = model.init_hidden(img_v1.shape[0])\n",
    "yinit_v1 = model(img_v1.transpose(1,0).transpose(0,2))\n",
    "print(\"Testing Accuracy: {0:.2f}\".format(accuracy(yinit_v1, lbl_v1) / float(img_v1.shape[0])))\n",
    "\n",
    "\n",
    "preds= torch.max(yinit_v1.data, 1)[1].cpu().numpy()\n",
    "true = lbl_v1.data.cpu().numpy()\n",
    "Conf_Labels = [0,1,2]\n",
    "conf_matrix = confusion_matrix(true, preds, Conf_Labels)\n",
    "fig = plt.figure()\n",
    "ax = fig.add_subplot(111)\n",
    "cax = ax.matshow(conf_matrix)\n",
    "plt.title('Confusion matrix')\n",
    "fig.colorbar(cax)\n",
    "ax.set_xticklabels([''] + Conf_Labels)\n",
    "ax.set_yticklabels([''] + Conf_Labels)\n",
    "plt.xlabel('Predicted')\n",
    "plt.ylabel('True')\n",
    "\n",
    "for i, j in itertools.product(range(conf_matrix.shape[0]), range(conf_matrix.shape[1])):\n",
    "        plt.text(j, i, format(conf_matrix[i, j]),horizontalalignment=\"center\",color=\"white\", fontsize=18)\n",
    "plt.show()\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 345,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(0, 1.111738681793213, 0.27011494252873564, 0.25806451612903225)\n",
      "Model File Saved !\n",
      "(50, 1.1617721319198608, 0.5689655172413793, 0.4838709677419355)\n",
      "Model File Saved !\n",
      "(100, 0.808106541633606, 0.632183908045977, 0.5161290322580645)\n",
      "Model File Saved !\n",
      "(150, 0.6556061506271362, 0.735632183908046, 0.6129032258064516)\n",
      "Model File Saved !\n",
      "(200, 0.45239004492759705, 0.7758620689655172, 0.6129032258064516)\n",
      "(250, 0.5738756060600281, 0.7758620689655172, 0.6129032258064516)\n",
      "(300, 0.5290581583976746, 0.7988505747126436, 0.5483870967741935)\n",
      "(350, 0.5624353885650635, 0.7931034482758621, 0.5483870967741935)\n",
      "(400, 0.3682890832424164, 0.8103448275862069, 0.4838709677419355)\n",
      "(450, 0.15060070157051086, 0.8160919540229885, 0.5161290322580645)\n",
      "(500, 0.38143157958984375, 0.8390804597701149, 0.5161290322580645)\n",
      "(550, 0.39412063360214233, 0.8448275862068966, 0.5161290322580645)\n",
      "(600, 0.22130770981311798, 0.8505747126436781, 0.5161290322580645)\n",
      "(650, 0.4254242479801178, 0.8563218390804598, 0.4838709677419355)\n",
      "(700, 0.4196566939353943, 0.8620689655172413, 0.5483870967741935)\n",
      "(750, 0.4564676582813263, 0.8793103448275862, 0.5161290322580645)\n",
      "(800, 0.17735497653484344, 0.896551724137931, 0.5483870967741935)\n",
      "(850, 0.2530496418476105, 0.8908045977011494, 0.5483870967741935)\n",
      "(900, 0.22624598443508148, 0.9022988505747126, 0.5483870967741935)\n",
      "(950, 0.08500552922487259, 0.8908045977011494, 0.4838709677419355)\n",
      "(1000, 0.14004693925380707, 0.9080459770114943, 0.5161290322580645)\n"
     ]
    }
   ],
   "source": [
    "mc = 0\n",
    "train_acc = []\n",
    "test_acc =  []\n",
    "acc = 0.0\n",
    "for epoch in range(num_epochs):\n",
    "    model.train()\n",
    "    mc = 0\n",
    "    for inp, out in data_loader:\n",
    "        model.zero_grad()\n",
    "        batch_size = inp.shape[0]\n",
    "        model.hidden = model.init_hidden(batch_size)\n",
    "        \n",
    "        inp = inp.transpose(1,0).transpose(0,2)\n",
    "        pred = model(inp)\n",
    "    \n",
    "        loss = loss_function(pred, out)\n",
    "        loss.backward()\n",
    "        \n",
    "        optimizer.step()\n",
    "        mc += ((torch.max(pred.data, 1)[1].long() == out.data).sum().item())\n",
    "        \n",
    "    if epoch % 50 == 0:\n",
    "        model.eval()\n",
    "        model.hidden = model.init_hidden(img_v1.shape[0])\n",
    "        yinit_v1 = model(img_v1.transpose(1,0).transpose(0,2))\n",
    "        test_acc.append(((accuracy(yinit_v1, lbl_v1) / float(img_v1.shape[0]))))\n",
    "        train_acc.append(mc/float(img.shape[0]))\n",
    "        print(epoch, loss.item(), mc/float(img.shape[0]), (accuracy(yinit_v1, lbl_v1)/ float(img_v1.shape[0])))\n",
    "        \n",
    "        if((accuracy(yinit_v1, lbl_v1)/ float(img_v1.shape[0])) > acc ):\n",
    "            torch.save(model.state_dict(), 'LSTMModel.pth')\n",
    "            acc = (accuracy(yinit_v1, lbl_v1)/ float(img_v1.shape[0]))\n",
    "            print('Model File Saved !')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 346,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "t = np.arange(num_epochs//50 + 1)\n",
    "plt.plot(t, train_acc)\n",
    "plt.plot(t, test_acc)\n",
    "plt.legend(['Train_Accuracy', 'Testing_Accuracy'])\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 347,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Testing Accuracy: 0.61\n"
     ]
    }
   ],
   "source": [
    "model = LSTM(3).cuda()\n",
    "model.load_state_dict(torch.load('LSTMModel.pth'))\n",
    "\n",
    "batch_size = img_v1.shape[0]\n",
    "model.hidden = model.init_hidden(batch_size)\n",
    "yinit_v1 = model(img_v1.transpose(1,0).transpose(0,2))\n",
    "print(\"Testing Accuracy: {0:.2f}\".format(accuracy(yinit_v1, lbl_v1) / float(img_v1.shape[0])))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 348,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "preds= torch.max(yinit_v1.data, 1)[1].cpu().numpy()\n",
    "true = lbl_v1.data.cpu().numpy()\n",
    "Conf_Labels = [0,1,2]\n",
    "conf_matrix = confusion_matrix(true, preds, Conf_Labels)\n",
    "fig = plt.figure()\n",
    "ax = fig.add_subplot(111)\n",
    "cax = ax.matshow(conf_matrix)\n",
    "plt.title('Confusion matrix')\n",
    "fig.colorbar(cax)\n",
    "ax.set_xticklabels([''] + Conf_Labels)\n",
    "ax.set_yticklabels([''] + Conf_Labels)\n",
    "plt.xlabel('Predicted')\n",
    "plt.ylabel('True')\n",
    "\n",
    "for i, j in itertools.product(range(conf_matrix.shape[0]), range(conf_matrix.shape[1])):\n",
    "        plt.text(j, i, format(conf_matrix[i, j]),horizontalalignment=\"center\",color=\"white\", fontsize=18)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 2",
   "language": "python",
   "name": "python2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.14"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}