show_2nd_order_similarity.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/javascript": [
       "IPython.notebook.set_autosave_interval(5000)"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Autosaving every 5 seconds\n"
     ]
    }
   ],
   "source": [
    "from qmvpa import utils, factor, rsa\n",
    "from data_loader_local import load_data\n",
    "import os\n",
    "import sys \n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt \n",
    "import seaborn as sns \n",
    "from dep.utils import subset_units\n",
    "from config import get_log_info\n",
    "from itertools import product\n",
    "\n",
    "%matplotlib inline \n",
    "%autosave 5\n",
    "sns.set(style = 'white', context='poster', font_scale=1, rc={\"lines.linewidth\": 2})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "log epochs [  0   1   2   3   5   8  13  22  36  60 100]\n",
      "layer selected [2, 10, 17, 25, 32, 40, 47, 55, 62]\n"
     ]
    }
   ],
   "source": [
    "# get roots \n",
    "log_root = '/tigress/qlu/logs/keras-resnet/log'\n",
    "plt_root = '/tigress/qlu/logs/keras-resnet/plots'\n",
    "pltdata_root = '/tigress/qlu/logs/keras-resnet/plots_data'\n",
    "\n",
    "# define data name, model name \n",
    "data_name = 'cifar100'\n",
    "model_name = 'resnet18'\n",
    "\n",
    "# load data \n",
    "_, _, _, _, _, labels, data_info = load_data(data_name)\n",
    "[n_classes, img_rows, img_cols, img_channels] = data_info\n",
    "n_test_egs = len(labels)\n",
    "\n",
    "# define params \n",
    "log_info_list = get_log_info(data_name, model_name)\n",
    "n_subjs, layer_selected, n_layers, log_epochs, n_log_epochs = log_info_list\n",
    "print('log epochs', log_epochs)\n",
    "print('layer selected', layer_selected)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "# collect precomputed RDMs\n",
    "wRSM_n_l2d = utils.list_2d(n_layers, n_log_epochs)\n",
    "iRSM_n_l2d = utils.list_2d(n_layers, n_log_epochs)\n",
    "wRSM_s_l2d = utils.list_2d(n_layers, n_log_epochs)\n",
    "iRSM_s_l2d = utils.list_2d(n_layers, n_log_epochs)\n",
    "\n",
    "# for each layer, look at the development \n",
    "for (i_l, i_e) in product(range(n_layers), range(n_log_epochs)):\n",
    "    l, e = layer_selected[i_l], log_epochs[i_e]\n",
    "\n",
    "    # plot dir and plot data dir \n",
    "    plt_dir = os.path.join(\n",
    "        plt_root, data_name, model_name, \n",
    "        'epoch_%.3d' % e, 'rsa')    \n",
    "    pltdata_dir = os.path.join(\n",
    "        pltdata_root, data_name, model_name, \n",
    "        'epoch_%.3d' % e, 'rsa')\n",
    "    # RSA and SRM result path\n",
    "    rsa_result_fname = 'wirsa_ns_l%.2d.npy' % (l)\n",
    "    rsa_result_path = os.path.join(pltdata_dir, rsa_result_fname)\n",
    "    srm_result_fname = 'srm_l%.2d.npy' % (l)\n",
    "    srm_result_path = os.path.join(pltdata_dir, srm_result_fname)\n",
    "    # unpack information\n",
    "    wRSM_n, iRSM_n, wRSM_s, iRSM_s = np.load(rsa_result_path)\n",
    "    wRSM_n_l2d[i_l][i_e] = wRSM_n\n",
    "    iRSM_n_l2d[i_l][i_e] = iRSM_n\n",
    "    wRSM_s_l2d[i_l][i_e] = wRSM_s\n",
    "    iRSM_s_l2d[i_l][i_e] = iRSM_s"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "\"\"\"\n",
    "collect correlations: \n",
    "inter-subject RSM at time t vs. final within-subject RSM\n",
    "\"\"\"\n",
    "R_vals = np.zeros((n_layers, n_log_epochs))\n",
    "P_vals = np.zeros((n_layers, n_log_epochs))\n",
    "\n",
    "for (i_l, i_e) in product(range(n_layers), range(n_log_epochs)):\n",
    "    l, e = layer_selected[i_l], log_epochs[i_e]\n",
    "    R_vals[i_l, i_e], P_vals[i_l, i_e] = rsa.correlate_2RSMs(\n",
    "        iRSM_s_l2d[i_l][i_e], wRSM_n_l2d[i_l][-1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "resnet18, cifar100\n",
      "\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 360x144 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "if model_name == 'conv':\n",
    "    n_layers = 3\n",
    "b_pals = sns.color_palette(\"Blues\", n_colors = n_layers)\n",
    "log_epochs_int = [int(log_epochs[i]) for i in range(len(log_epochs))]\n",
    "\n",
    "\n",
    "f, ax = plt.subplots(1,1, figsize = (5, 2))\n",
    "\n",
    "# plot a heatmap to get the color bar for neural network depth \n",
    "temp_plt = sns.heatmap(\n",
    "    R_vals, cmap=sns.color_palette(\"Blues\", n_colors = n_layers),\n",
    "    cbar_kws={'label': 'Layer depth', 'ticks' : [0, n_layers]},\n",
    "    ax = ax)\n",
    "ax.cla()\n",
    "\n",
    "for i_l in range(n_layers):\n",
    "    ax.plot(log_epochs_int, R_vals[i_l,:], color = b_pals[i_l], marker = '.')\n",
    "\n",
    "title_text = \"\"\"\n",
    "%s, %s\n",
    "\"\"\" % (model_name, data_name)\n",
    "# title_text = \"\"\"\n",
    "# 2nd order correlation\n",
    "# inter-subject RSM vs. final within-subject RSM\n",
    "# model: %s, data: %s\n",
    "# \"\"\" % (model_name, data_name)\n",
    "ax.set_title(title_text)\n",
    "ax.set_xlabel('Training epochs')\n",
    "ax.set_ylabel('Linear corr.')\n",
    "\n",
    "ax.set_ylim([0, 1.05]);\n",
    "ax.set_xlim([0, log_epochs_int[-1]+1]);\n",
    "sns.despine()\n",
    "\n",
    "print(title_text)\n",
    "\n",
    "temp_path = '/tigress/qlu/logs/temp/'\n",
    "fname = 'deve_%s_%s_s.png' % (model_name, data_name)\n",
    "f.savefig(os.path.join(temp_path, fname),  bbox_inches='tight', dpi = 1000)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "\"\"\"\n",
    "collect correlations: \n",
    "inter-subject RSM at time t vs. within-subject RSM at time t\n",
    "\"\"\"\n",
    "R_vals = np.zeros((n_layers, n_log_epochs))\n",
    "P_vals = np.zeros((n_layers, n_log_epochs))\n",
    "\n",
    "for (i_l, i_e) in product(range(n_layers), range(n_log_epochs)):\n",
    "    l, e = layer_selected[i_l], log_epochs[i_e]\n",
    "    R_vals[i_l, i_e], P_vals[i_l, i_e] = rsa.correlate_2RSMs(\n",
    "        iRSM_s_l2d[i_l][i_e], wRSM_n_l2d[i_l][i_e])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 360x144 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "if model_name == 'conv':\n",
    "    n_layers = 3\n",
    "b_pals = sns.color_palette(\"Blues\", n_colors = n_layers)\n",
    "log_epochs_int = [int(log_epochs[i]) for i in range(len(log_epochs))]\n",
    "\n",
    "f, ax = plt.subplots(1,1, figsize = (5, 2))\n",
    "temp_plt = sns.heatmap(\n",
    "    R_vals, cmap=sns.color_palette(\"Blues\", n_colors = n_layers),\n",
    "    cbar_kws={'label': 'Layer depth', 'ticks' : [0, n_layers]},\n",
    "    ax = ax)\n",
    "ax.cla()\n",
    "\n",
    "for i_l in range(n_layers):\n",
    "    ax.plot(log_epochs_int, R_vals[i_l,:], color = b_pals[i_l], marker = '.')\n",
    "\n",
    "title_text = \"\"\"\n",
    "%s, %s\n",
    "\"\"\" % (model_name, data_name)\n",
    "ax.set_title(title_text)\n",
    "ax.set_xlabel('Training epochs')\n",
    "ax.set_ylabel('Linear corr.')\n",
    "\n",
    "ax.set_ylim([0, 1.05]);\n",
    "ax.set_xlim([0, log_epochs_int[-1]+1]);\n",
    "sns.despine()"
   ]
  },
  {
   "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": "3.6.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}