generate_LSP_vectors.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Table of Contents:\n",
    "1. Loss Weight Scheduler plots\n",
    "2. LSP generator"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 864x432 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "from matplotlib import rcParams\n",
    "rcParams['font.weight'] = 'bold'\n",
    "\n",
    "lambda_og = lambda epoch: 1 if epoch <= 400 else 1.95 - 0.95/400 * epoch if epoch <= 800 else 0.05\n",
    "lambda_pseudo_train = lambda epoch: 0 if epoch <= 400 else -1 + epoch/400 if epoch <= 800 else 1\n",
    "lambda_pseudo_random = lambda epoch: 0 if epoch <= 800 else -2 + epoch/400 if epoch <= 1200 else 1\n",
    "\n",
    "x = range(2000)\n",
    "og_values = [lambda_og(v) for v in x]\n",
    "pseudo_train_values = [lambda_pseudo_train(v) for v in x]\n",
    "pseudo_random_values = [lambda_pseudo_random(v) for v in x]\n",
    "\n",
    "sns.set_style(\"white\")  # Set plot style\n",
    "plt.figure(figsize=(12, 6))  # Set figure size\n",
    "\n",
    "# Plot the lines\n",
    "linewidth = 3\n",
    "og_line = sns.lineplot(x=x, y=og_values, label=\"Original Labels\", linewidth=linewidth) #, color=\"#DBEDFF\")\n",
    "pseudo_train_line = sns.lineplot(x=x, y=pseudo_train_values, label=\"Pseudo Labels (Train)\", linewidth=linewidth) #,color=\"#FFEBD6\")\n",
    "pseudo_random_line = sns.lineplot(x=x, y=pseudo_random_values, label=\"Pseudo Labels (Random)\", linewidth=linewidth) #, color=\"#E8F3E7\")\n",
    "\n",
    "fontsize = 20\n",
    "plt.annotate(\"0.05\" ,xy=(1000, pseudo_train_line.lines[0].get_ydata()[1000]), xytext=(1850, pseudo_train_line.lines[0].get_ydata()[1000] + 0.03),\n",
    "             arrowprops=dict(arrowstyle=\"->\"), fontsize=20)\n",
    "plt.xticks(range(0, 2001, 400), fontsize=fontsize, fontweight=\"bold\")\n",
    "plt.yticks(fontsize=fontsize, fontweight=\"bold\")\n",
    "plt.xlabel(\"Epochs\", fontsize=fontsize, fontweight=\"bold\")\n",
    "plt.ylabel(\"Weight\", fontsize=fontsize, fontweight=\"bold\")\n",
    "plt.title(\"Weight Schedules for Labels\", fontsize=fontsize, fontweight=\"bold\")\n",
    "plt.legend(fontsize=15)\n",
    "plt.savefig(\"weight_scheduler.png\", dpi=300)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "2. LSP vector generator"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Using GPU: NVIDIA RTX A5000\n",
      "Found local copy...\n",
      "Found local copy...\n",
      "Found local copy...\n",
      "Found saved processed KG... Loading...\n",
      "Splits detected... Loading splits....\n",
      "Creating DGL graph....\n",
      "Done!\n"
     ]
    }
   ],
   "source": [
    "from txgnn import TxGNN, TxEval, TxData, NewDataHandler\n",
    "from txgnn.utils import *\n",
    "from tqdm import tqdm\n",
    "import time\n",
    "import argparse\n",
    "\n",
    "split = \"complex_disease\"\n",
    "seed = 45\n",
    "size = 130\n",
    "\n",
    "TxData1 = TxData(data_folder_path = './data/')\n",
    "TxData1.prepare_split(split = split, seed = seed, no_kg = False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "TxGNN1 = TxGNN(\n",
    "        data = TxData1, \n",
    "        weight_bias_track = False,\n",
    "        proj_name = 'TxGNN',\n",
    "        exp_name = 'TxGNN'\n",
    "    )\n",
    "class fake_argparse():\n",
    "    pass\n",
    "pseudo_args = fake_argparse()\n",
    "pseudo_args.all_layer_LSP = False\n",
    "pseudo_args.pretrain_scores_dict = None\n",
    "pseudo_args.on_the_fly_KD = False\n",
    "TxGNN1.model_initialize(n_hid = size, \n",
    "                        n_inp = size, \n",
    "                        n_out = size, \n",
    "                        proto = True,\n",
    "                        proto_num = 3,\n",
    "                        attention = False,\n",
    "                        sim_measure = 'all_nodes_profile',\n",
    "                        bert_measure = 'disease_name',\n",
    "                        agg_measure = 'rarity',\n",
    "                        num_walks = 200,\n",
    "                        walk_mode = 'bit',\n",
    "                        path_length = 2,\n",
    "                        LSP_size=\"partial\",\n",
    "                        args=pseudo_args\n",
    "                        )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loaded a trained model from path: pretrained_models/130_e2000/45. Could be fully trained or only the pretrain phase\n",
      "Accuracy for Positive train graph is 0.9341320991516113\n",
      "Accuracy for Negative train graph is 0.8720200657844543\n",
      "Accuracy for Positive val graph is 0.47529128193855286\n",
      "Accuracy for Negative val graph is 0.8563680648803711\n",
      "Accuracy for Positive test graph is 0.5615196824073792\n",
      "Accuracy for Negative test graph is 0.8441965579986572\n",
      "Validation Loss 0.9950,  Validation Micro AUROC 0.7455 Validation Micro AUPRC 0.7546 Validation Macro AUROC 0.7863 Validation Macro AUPRC 0.7921\n",
      "----- AUROC Performance in Each Relation -----\n",
      "('drug', 'contraindication', 'disease'): 0.7175645061027708\n",
      "('drug', 'indication', 'disease'): 0.8550084192757323\n",
      "----- AUPRC Performance in Each Relation -----\n",
      "('drug', 'contraindication', 'disease'): 0.720437972818933\n",
      "('drug', 'indication', 'disease'): 0.8636847093658915\n",
      "----------------------------------------------\n",
      "Testing Loss 0.8000 Testing Micro AUROC 0.7877 Testing Micro AUPRC 0.7638 Testing Macro AUROC 0.8252 Testing Macro AUPRC 0.8033\n",
      "----- AUROC Performance in Each Relation -----\n",
      "('drug', 'contraindication', 'disease'): 0.7523534109453822\n",
      "('drug', 'indication', 'disease'): 0.89794697156476\n",
      "----- AUPRC Performance in Each Relation -----\n",
      "('drug', 'contraindication', 'disease'): 0.722917653597591\n",
      "('drug', 'indication', 'disease'): 0.8837328801236818\n",
      "----------------------------------------------\n"
     ]
    }
   ],
   "source": [
    "save_dir = f'pretrained_models/{size}_e2000/{seed}'\n",
    "TxGNN1.load_pretrained(save_dir)\n",
    "\n",
    "G = TxGNN1.G.to(TxGNN1.device)\n",
    "best_G = TxGNN1.best_G.to(TxGNN1.device)\n",
    "best_model = TxGNN1.best_model.to(TxGNN1.device)\n",
    "best_model.eval()\n",
    "\n",
    "## Print out accuracies\n",
    "neg_sampler = Full_Graph_NegSampler(TxGNN1.G, 1, 'fix_dst', TxGNN1.device)\n",
    "g_train_neg = neg_sampler(TxGNN1.G)\n",
    "g_valid_pos, g_valid_neg = TxGNN1.g_valid_pos, TxGNN1.g_valid_neg\n",
    "g_test_pos, g_test_neg = TxGNN1.g_test_pos, TxGNN1.g_test_neg\n",
    "## arguments: (model, G, train_pos, train_neg, val_pos, val_neg, test_pos, test_neg, device=None)\n",
    "accuracies_dict = evaluate_accuracy_per_split(best_model, best_G, g_train_neg, g_valid_pos, g_valid_neg, g_test_pos, g_test_neg)\n",
    "for k, v in accuracies_dict.items():\n",
    "    print(f\"Accuracy for Positive {k} graph is {v[0].item()}\")\n",
    "    print(f\"Accuracy for Negative {k} graph is {v[1].item()}\")\n",
    "print_val_test_auprc(best_model, g_valid_pos, g_valid_neg, g_test_pos, g_test_neg, best_G, TxGNN1.dd_etypes, TxGNN1.device)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "time it took to create the sparse tensor: 2.2320377826690674\n",
      "time it took to create the sparse tensor: 2.2454607486724854\n"
     ]
    }
   ],
   "source": [
    "best_G = TxGNN1.G.to(TxGNN1.device)\n",
    "best_model = TxGNN1.best_model.to(TxGNN1.device)\n",
    "inter_h, h, beta_kl_loss, distmult = best_model(best_G, pretrain_mode = False, mode = 'train', return_h_and_kl=True, return_all_layer_h=True)\n",
    "LS_preds = []\n",
    "for h, sigma in zip([inter_h, h], [0.2, 3]):\n",
    "    _, _, LS = distmult(best_G, best_G, h, mode=None, pretrain_mode=False, LSP=\"RBF\", LSP_size=\"partial\", sigma=sigma)\n",
    "    LS_prob = []\n",
    "    for i, v in enumerate(LS):\n",
    "        softmaxed = F.softmax(v)\n",
    "        LS_prob.append(softmaxed)\n",
    "    LS_preds.append(LS_prob)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "torch.save(LS_preds[1], f\"LSP_{size}_{seed}_partial_RBF.pt\") ## saves 1 layer LSP\n",
    "torch.save(LS_preds, f\"LSP_{size}_double_{seed}_partial_RBF.pt\") ## saves both layer LSP"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "mykernel",
   "language": "python",
   "name": "mykernel"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.9.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}