hcp_data/lstm-book.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pickle\n",
    "import numpy as np\n",
    "import torch\n",
    "import random\n",
    "from matplotlib import pyplot as plt\n",
    "from torch.utils.data import TensorDataset, DataLoader\n",
    "\n",
    "from lstm_data_prep import prep, numpy_prep\n",
    "from rnn_perm_test import iterateLSTM, create_permutation_3d\n",
    "from lstm import test_model_lstm, initialize_lstm"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "c:\\Users\\Anna&Megan\\Documents\\GitHub\\srp-lab-notebook\\hcp_data\\lstm_data_prep.py:59: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray.\n",
      "  X_padded = paddingArr(np.array(X_arr))\n"
     ]
    }
   ],
   "source": [
    "device = \"cuda\" if torch.cuda.is_available() else \"cpu\"\n",
    "with open('HCP_movie_watching.pkl','rb') as f:\n",
    "    TS = pickle.load(f)\n",
    "\n",
    "input_size = 300\n",
    "hidden_size = 32\n",
    "seq_len = 90\n",
    "class_num = 15\n",
    "\n",
    "train_loader, test_loader = prep(TS)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "lstmModel = initialize_lstm(input_size, hidden_size, class_num, seq_len)\n",
    "train_loss = lstmModel.train(train_loader, 100)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "lstmModel = initialize_lstm(input_size, hidden_size, seq_len)\n",
    "\n",
    "lstm = lstmModel.model\n",
    "lstmLoss = lstmModel.loss_fn\n",
    "lstmOptim = lstmModel.optim\n",
    "\n",
    "check = torch.load(\"lstm-model.pt\")\n",
    "lstm.load_state_dict(check[\"lstm\"])\n",
    "lstmOptim.load_state_dict(check[\"lstm_optimizer\"])\n",
    "\n",
    "lstm.eval()\n",
    "\n",
    "lstm_accuracy = lstmModel.test_lstm(test_loader)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Run permutation test on model\n",
    "# permutation_lstm_accuracy = iterateLSTM(lstmModel, lstmLoss, TS, num_samples=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "def generate_random_features(X_lens, num_batches, num_seq = 90, num_features = 300, pad=-100.):\n",
    "    X_random = []\n",
    "    for i in range(num_batches):\n",
    "        X_batch = np.random.normal(size=(X_lens[i], num_features))\n",
    "        if X_lens[i] < num_seq:\n",
    "            X_pad = np.array([[-100.]*300]*(num_seq - X_lens[i]))\n",
    "            X_batch = np.append(X_batch, X_pad, axis=0)\n",
    "        X_random.append(X_batch)\n",
    "    X_random = np.array(X_random)\n",
    "    return X_random\n",
    "\n",
    "def test_random_features(model, dictionary, seq_len = 90, num_samples = 1, batch_size = 32):\n",
    "    _, _, X, y = numpy_prep(dictionary)\n",
    "    X_lens = find_lens(X)\n",
    "    random_features_acc = []\n",
    "    for i in range(num_samples):\n",
    "        X_random = generate_random_features(X_lens, X.shape[0])\n",
    "        X_random_data = TensorDataset(torch.from_numpy(X_random).float(), torch.from_numpy(y).float())\n",
    "        X_random_loader = DataLoader(X_random_data, shuffle=True, batch_size=batch_size)\n",
    "        sample_acc = model.test_lstm(X_random_loader)\n",
    "        random_features_acc.append(sample_acc)\n",
    "    return random_features_acc"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "def find_lens(X):\n",
    "    X_lens = []\n",
    "    for batch in X:\n",
    "        count = 0\n",
    "        for time in batch:\n",
    "            if time[0] == -100.:\n",
    "                break\n",
    "            else:\n",
    "                count += 1\n",
    "        X_lens.append(count)\n",
    "    return X_lens"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Stored 'permutation_lstm_accuracy' (ndarray)\n"
     ]
    }
   ],
   "source": [
    "%store permutation_lstm_accuracy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Compare accuracies\n",
    "xAx = [i for i in range(0,90)]\n",
    "plt.plot(xAx, lstm_accuracy, label=\"lstm\")\n",
    "plt.xlabel(\"Time (s)\")\n",
    "plt.ylabel(\"Accuracy\")\n",
    "plt.ylim(0,1)\n",
    "plt.xlim(0,90)\n",
    "plt.title(\"Time-varying Classification Accuracy\")\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Stored 'lstm_accuracy' (ndarray)\n"
     ]
    }
   ],
   "source": [
    "%store lstm_accuracy"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3.10.5 64-bit",
   "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": "3.10.5"
  },
  "orig_nbformat": 4,
  "vscode": {
   "interpreter": {
    "hash": "5d8d8f94dc29cf6517d9b951f40e6c965bcb2efc4a5d0d869ef8b359fa785960"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}