Compare Hungarian algorithm implementations.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Code comparing align_clusters_groups and scipy.optimize.linear_sum_assignment\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-22T04:09:11.753045Z",
     "start_time": "2019-06-22T04:09:11.449902Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Max Correlations using align_clusters_groups\n",
      "Iter 0 0.263\n",
      "Iter 1 0.291\n",
      "Iter 2 0.235\n",
      "Iter 3 0.267\n",
      "Iter 4 0.255\n",
      "Iter 5 0.29\n",
      "Iter 6 0.278\n",
      "Iter 7 0.281\n",
      "Iter 8 0.266\n",
      "Iter 9 0.268\n",
      "\n",
      "\n",
      "Max Correlations using scipy.optimize.linear_sum_assignment\n",
      "Iter 0 0.312\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/jinhyuncheong/Dropbox/Packages/fnl_tools/fnl_tools/stats.py:190: FutureWarning: '.reindex_axis' is deprecated and will be removed in a future version. Use '.reindex' instead.\n",
      "  group2_new = group2_new.reindex_axis(sorted(group2_new.columns), axis=1)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Iter 1 0.312\n",
      "Iter 2 0.312\n",
      "Iter 3 0.312\n",
      "Iter 4 0.312\n",
      "Iter 5 0.312\n",
      "Iter 6 0.312\n",
      "Iter 7 0.312\n",
      "Iter 8 0.312\n",
      "Iter 9 0.312\n"
     ]
    }
   ],
   "source": [
    "cols = 20\n",
    "times = 30\n",
    "iters = 10\n",
    "import numpy as np, pandas as pd\n",
    "np.random.seed(100)\n",
    "# s1 & s2 are the two groups of states to be aligned together.\n",
    "s1 = pd.DataFrame(np.random.rand(times,cols),columns=['s1_'+str(i) for i in range(cols)])\n",
    "s2 = pd.DataFrame(np.random.rand(times,cols),columns=['s2_'+str(i) for i in range(cols)])\n",
    "\n",
    "# Compute Max Correlation with align_clusters_groups\n",
    "print(\"Max Correlations using align_clusters_groups\")\n",
    "from fnl_tools.stats import align_clusters_groups\n",
    "for i in range(iters):\n",
    "    # Run Align Clusters with s1 and column-shuffled s2\n",
    "    (group2_new, remapped_columns) = align_clusters_groups(s1, s2[np.random.permutation(s2.columns)])\n",
    "    # Horizontal stack s1 and reordered s2\n",
    "    srms = pd.concat([s1,group2_new], axis=1)\n",
    "    srms_corr = srms.corr()\n",
    "    # Max correlation is the diagonal of the upper right quadrant. \n",
    "    print(f\"Iter {i}\", np.round(np.mean(np.diag(srms_corr.iloc[:20, 20:])),3))\n",
    "print('\\n')\n",
    "# Compute Max Correlation with scipy.optimize.linear_sum_assignment.\n",
    "print(\"Max Correlations using scipy.optimize.linear_sum_assignment\")\n",
    "import scipy\n",
    "for i in range(iters):\n",
    "    # Horizontal stack s1 and s2 \n",
    "    srms = pd.concat([s1,s2[np.random.permutation(s2.columns)]], axis=1)\n",
    "    # Get the \"cost\" matrix by getting correlations of upper right quadrant.\n",
    "    srms_corr = srms.corr().iloc[:20, 20:]\n",
    "    # Align using Hungarian algorithm with correlation matrix as \"cost\" by inverting the correlation. \n",
    "    row_ind, col_ind = scipy.optimize.linear_sum_assignment(-1*srms_corr)\n",
    "    # Max correlation is the diagonal of the upper right quadrant.\n",
    "    print(f\"Iter {i}\",np.round(np.mean(np.diag(srms_corr.iloc[row_ind,col_ind])),3))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "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": "3.6.5"
  },
  "toc": {
   "nav_menu": {
    "height": "45px",
    "width": "252px"
   },
   "navigate_menu": true,
   "number_sections": true,
   "sideBar": true,
   "threshold": 4,
   "toc_cell": false,
   "toc_section_display": "block",
   "toc_window_display": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}