Parallel K-Modes and K-Prototypes

test.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import time\n",
    "\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "\n",
    "from kmodes.kmodes import KModes\n",
    "from kmodes.kprototypes import KPrototypes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "data = ...\n",
    "cat_cols = data.x_data.select_dtypes('category').columns\n",
    "cat_index = [data.x_data.columns.get_loc(column)\n",
    "            for column, dtype in data.x_data.dtypes.items()\n",
    "            if pd.api.types.is_categorical_dtype(dtype)]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Finished with 1 jobs in 14.94139575958252 s.\n",
      "Finished with 2 jobs in 7.5124900341033936 s.\n",
      "Finished with 3 jobs in 5.480122327804565 s.\n",
      "Finished with 4 jobs in 4.830286026000977 s.\n",
      "Finished with 5 jobs in 5.453425884246826 s.\n",
      "Finished with 6 jobs in 5.0629661083221436 s.\n",
      "Finished with 7 jobs in 5.260700702667236 s.\n",
      "Finished with 8 jobs in 6.280949831008911 s.\n"
     ]
    }
   ],
   "source": [
    "labels = []\n",
    "for n_jobs in range(1, 9):\n",
    "    np.random.seed(42)\n",
    "    kproto = KPrototypes(20, n_jobs=n_jobs, random_state=42)\n",
    "    start = time.time()\n",
    "    kproto.fit(data.x_data.sample(n=1000, random_state=42), data.x_data, categorical=cat_index)\n",
    "    print(\"Finished with {} jobs in {} s.\".format(n_jobs, time.time() - start))\n",
    "    labels.append(kproto.labels_)\n",
    "    \n",
    "    if n_jobs > 1:\n",
    "        assert np.all(np.equal(labels[n_jobs-2], labels[n_jobs-1]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Finished with 1 jobs in 22.188632011413574 s.\n",
      "Finished with 2 jobs in 26.93237280845642 s.\n",
      "Finished with 3 jobs in 26.965989112854004 s.\n",
      "Finished with 4 jobs in 26.93553900718689 s.\n",
      "Finished with 5 jobs in 27.35899806022644 s.\n",
      "Finished with 6 jobs in 28.72810435295105 s.\n",
      "Finished with 7 jobs in 27.51678490638733 s.\n",
      "Finished with 8 jobs in 27.7084379196167 s.\n"
     ]
    }
   ],
   "source": [
    "labels = []\n",
    "for n_jobs in range(1, 9):\n",
    "    np.random.seed(42)\n",
    "    kmodes = KModes(20, n_jobs=n_jobs, random_state=42)\n",
    "    start = time.time()\n",
    "    kmodes.fit(data.x_data[cat_cols].sample(n=100000, random_state=42), data.x_data)\n",
    "    print(\"Finished with {} jobs in {} s.\".format(n_jobs, time.time() - start))\n",
    "    labels.append(kmodes.labels_)\n",
    "    \n",
    "    if n_jobs > 1:\n",
    "        assert np.all(np.equal(labels[n_jobs-2], labels[n_jobs-1]))"
   ]
  }
 ],
 "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"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}

Thanks @rphes for the parallel execution in Kprototypes