generate-train-test-sample-from-hr4-catalog-pyspark-wip.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# PySpark Job"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Trim `hrdata.parquet.snappy` for a training sample"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "# generate edges \n",
    "import sys\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "from scipy.spatial import cKDTree\n",
    "import gc\n",
    "\n",
    "# plot settings\n",
    "plt.rc('font', family='serif') \n",
    "plt.rc('font', serif='Times New Roman') \n",
    "plt.rcParams.update({'font.size': 16})\n",
    "plt.rcParams['mathtext.fontset'] = 'stix'"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Let's determine the sizes of traing/testing samples"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "29791\n"
     ]
    }
   ],
   "source": [
    "print 31**3"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2883\n"
     ]
    }
   ],
   "source": [
    "print 31**2*3"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "27000\n"
     ]
    }
   ],
   "source": [
    "print 30**3"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "numsect = 31**3"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "> 20000 training sample and 7000 testing sample "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Read `hrdata.parquet.snappy` to a pandas/dataframe, and trim it!"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pyarrow.parquet as pq"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 3 µs, sys: 3 µs, total: 6 µs\n",
      "Wall time: 7.15 µs\n"
     ]
    }
   ],
   "source": [
    "%time\n",
    "hrdf = pq.read_table('hrdata.parquet.snappy').to_pandas()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>haloid</th>\n",
       "      <th>px</th>\n",
       "      <th>py</th>\n",
       "      <th>pz</th>\n",
       "      <th>halomass</th>\n",
       "      <th>isect</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>2164471</td>\n",
       "      <td>113.276787</td>\n",
       "      <td>136.558777</td>\n",
       "      <td>158.994751</td>\n",
       "      <td>7.075400e+14</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>341850</td>\n",
       "      <td>148.616104</td>\n",
       "      <td>193.995865</td>\n",
       "      <td>172.711838</td>\n",
       "      <td>6.187440e+14</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2654219</td>\n",
       "      <td>156.185623</td>\n",
       "      <td>115.296417</td>\n",
       "      <td>168.603867</td>\n",
       "      <td>5.634040e+14</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>1339587</td>\n",
       "      <td>188.427338</td>\n",
       "      <td>189.256927</td>\n",
       "      <td>168.730972</td>\n",
       "      <td>3.790270e+14</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>24784</td>\n",
       "      <td>145.331253</td>\n",
       "      <td>188.979660</td>\n",
       "      <td>175.401337</td>\n",
       "      <td>3.555020e+14</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    haloid          px          py          pz      halomass  isect\n",
       "0  2164471  113.276787  136.558777  158.994751  7.075400e+14      0\n",
       "1   341850  148.616104  193.995865  172.711838  6.187440e+14      0\n",
       "2  2654219  156.185623  115.296417  168.603867  5.634040e+14      0\n",
       "3  1339587  188.427338  189.256927  168.730972  3.790270e+14      0\n",
       "4    24784  145.331253  188.979660  175.401337  3.555020e+14      0"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "hrdf.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "Int64Index: 54000000 entries, 0 to 53999999\n",
      "Data columns (total 6 columns):\n",
      "haloid      int32\n",
      "px          float32\n",
      "py          float32\n",
      "pz          float32\n",
      "halomass    float32\n",
      "isect       int64\n",
      "dtypes: float32(4), int32(1), int64(1)\n",
      "memory usage: 1.8 GB\n"
     ]
    }
   ],
   "source": [
    "hrdf.info()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "54000000"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(hrdf.index)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[0, 26999]"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "[hrdf['isect'].min(),hrdf['isect'].max()]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([    0,     1,     2, ..., 26997, 26998, 26999])"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "hrdf.isect.unique()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Grouping `hrdf` based on the `isect`, and test the `grouped`  made by `groupby('isect')` "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "27000"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(hrdf.groupby('isect'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Int64Index([4000000, 4000001, 4000002, 4000003, 4000004, 4000005, 4000006,\n",
       "            4000007, 4000008, 4000009],\n",
       "           dtype='int64')"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "hrdf.groupby('isect').groups[2000][:10]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>haloid</th>\n",
       "      <th>px</th>\n",
       "      <th>py</th>\n",
       "      <th>pz</th>\n",
       "      <th>halomass</th>\n",
       "      <th>isect</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>4000000</th>\n",
       "      <td>134359</td>\n",
       "      <td>335.164948</td>\n",
       "      <td>742.002258</td>\n",
       "      <td>2128.719238</td>\n",
       "      <td>4.283600e+14</td>\n",
       "      <td>2000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4000001</th>\n",
       "      <td>2245720</td>\n",
       "      <td>320.469086</td>\n",
       "      <td>754.580933</td>\n",
       "      <td>2131.292969</td>\n",
       "      <td>3.453720e+14</td>\n",
       "      <td>2000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4000002</th>\n",
       "      <td>480008</td>\n",
       "      <td>342.707184</td>\n",
       "      <td>787.741760</td>\n",
       "      <td>2101.839600</td>\n",
       "      <td>1.780080e+14</td>\n",
       "      <td>2000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4000003</th>\n",
       "      <td>134762</td>\n",
       "      <td>326.419830</td>\n",
       "      <td>713.458313</td>\n",
       "      <td>2188.218262</td>\n",
       "      <td>1.191590e+14</td>\n",
       "      <td>2000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4000004</th>\n",
       "      <td>3193581</td>\n",
       "      <td>315.241241</td>\n",
       "      <td>712.992493</td>\n",
       "      <td>2151.678467</td>\n",
       "      <td>1.149920e+14</td>\n",
       "      <td>2000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4000005</th>\n",
       "      <td>6551121</td>\n",
       "      <td>323.842133</td>\n",
       "      <td>745.397156</td>\n",
       "      <td>2142.352051</td>\n",
       "      <td>1.144140e+14</td>\n",
       "      <td>2000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4000006</th>\n",
       "      <td>6330114</td>\n",
       "      <td>300.389343</td>\n",
       "      <td>705.541321</td>\n",
       "      <td>2109.851807</td>\n",
       "      <td>1.141620e+14</td>\n",
       "      <td>2000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4000007</th>\n",
       "      <td>440544</td>\n",
       "      <td>388.964355</td>\n",
       "      <td>749.215515</td>\n",
       "      <td>2196.093262</td>\n",
       "      <td>1.122040e+14</td>\n",
       "      <td>2000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4000008</th>\n",
       "      <td>8168306</td>\n",
       "      <td>389.101257</td>\n",
       "      <td>703.775879</td>\n",
       "      <td>2165.708740</td>\n",
       "      <td>1.102020e+14</td>\n",
       "      <td>2000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4000009</th>\n",
       "      <td>249409</td>\n",
       "      <td>357.410156</td>\n",
       "      <td>746.634705</td>\n",
       "      <td>2174.904541</td>\n",
       "      <td>1.027600e+14</td>\n",
       "      <td>2000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "          haloid          px          py           pz      halomass  isect\n",
       "4000000   134359  335.164948  742.002258  2128.719238  4.283600e+14   2000\n",
       "4000001  2245720  320.469086  754.580933  2131.292969  3.453720e+14   2000\n",
       "4000002   480008  342.707184  787.741760  2101.839600  1.780080e+14   2000\n",
       "4000003   134762  326.419830  713.458313  2188.218262  1.191590e+14   2000\n",
       "4000004  3193581  315.241241  712.992493  2151.678467  1.149920e+14   2000\n",
       "4000005  6551121  323.842133  745.397156  2142.352051  1.144140e+14   2000\n",
       "4000006  6330114  300.389343  705.541321  2109.851807  1.141620e+14   2000\n",
       "4000007   440544  388.964355  749.215515  2196.093262  1.122040e+14   2000\n",
       "4000008  8168306  389.101257  703.775879  2165.708740  1.102020e+14   2000\n",
       "4000009   249409  357.410156  746.634705  2174.904541  1.027600e+14   2000"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "hrdf[hrdf.isect == 2000][:10]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "grouped = hrdf.groupby('isect')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>haloid</th>\n",
       "      <th>px</th>\n",
       "      <th>py</th>\n",
       "      <th>pz</th>\n",
       "      <th>halomass</th>\n",
       "      <th>isect</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>4000000</th>\n",
       "      <td>134359</td>\n",
       "      <td>335.164948</td>\n",
       "      <td>742.002258</td>\n",
       "      <td>2128.719238</td>\n",
       "      <td>4.283600e+14</td>\n",
       "      <td>2000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4000001</th>\n",
       "      <td>2245720</td>\n",
       "      <td>320.469086</td>\n",
       "      <td>754.580933</td>\n",
       "      <td>2131.292969</td>\n",
       "      <td>3.453720e+14</td>\n",
       "      <td>2000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4000002</th>\n",
       "      <td>480008</td>\n",
       "      <td>342.707184</td>\n",
       "      <td>787.741760</td>\n",
       "      <td>2101.839600</td>\n",
       "      <td>1.780080e+14</td>\n",
       "      <td>2000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4000003</th>\n",
       "      <td>134762</td>\n",
       "      <td>326.419830</td>\n",
       "      <td>713.458313</td>\n",
       "      <td>2188.218262</td>\n",
       "      <td>1.191590e+14</td>\n",
       "      <td>2000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4000004</th>\n",
       "      <td>3193581</td>\n",
       "      <td>315.241241</td>\n",
       "      <td>712.992493</td>\n",
       "      <td>2151.678467</td>\n",
       "      <td>1.149920e+14</td>\n",
       "      <td>2000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "          haloid          px          py           pz      halomass  isect\n",
       "4000000   134359  335.164948  742.002258  2128.719238  4.283600e+14   2000\n",
       "4000001  2245720  320.469086  754.580933  2131.292969  3.453720e+14   2000\n",
       "4000002   480008  342.707184  787.741760  2101.839600  1.780080e+14   2000\n",
       "4000003   134762  326.419830  713.458313  2188.218262  1.191590e+14   2000\n",
       "4000004  3193581  315.241241  712.992493  2151.678467  1.149920e+14   2000"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "grouped.get_group(2000).head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "tmp = grouped.get_group(2000)[:5]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>haloid</th>\n",
       "      <th>px</th>\n",
       "      <th>py</th>\n",
       "      <th>pz</th>\n",
       "      <th>halomass</th>\n",
       "      <th>isect</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>4000000</th>\n",
       "      <td>134359</td>\n",
       "      <td>335.164948</td>\n",
       "      <td>742.002258</td>\n",
       "      <td>2128.719238</td>\n",
       "      <td>4.283600e+14</td>\n",
       "      <td>2000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4000001</th>\n",
       "      <td>2245720</td>\n",
       "      <td>320.469086</td>\n",
       "      <td>754.580933</td>\n",
       "      <td>2131.292969</td>\n",
       "      <td>3.453720e+14</td>\n",
       "      <td>2000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4000002</th>\n",
       "      <td>480008</td>\n",
       "      <td>342.707184</td>\n",
       "      <td>787.741760</td>\n",
       "      <td>2101.839600</td>\n",
       "      <td>1.780080e+14</td>\n",
       "      <td>2000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4000003</th>\n",
       "      <td>134762</td>\n",
       "      <td>326.419830</td>\n",
       "      <td>713.458313</td>\n",
       "      <td>2188.218262</td>\n",
       "      <td>1.191590e+14</td>\n",
       "      <td>2000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4000004</th>\n",
       "      <td>3193581</td>\n",
       "      <td>315.241241</td>\n",
       "      <td>712.992493</td>\n",
       "      <td>2151.678467</td>\n",
       "      <td>1.149920e+14</td>\n",
       "      <td>2000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "          haloid          px          py           pz      halomass  isect\n",
       "4000000   134359  335.164948  742.002258  2128.719238  4.283600e+14   2000\n",
       "4000001  2245720  320.469086  754.580933  2131.292969  3.453720e+14   2000\n",
       "4000002   480008  342.707184  787.741760  2101.839600  1.780080e+14   2000\n",
       "4000003   134762  326.419830  713.458313  2188.218262  1.191590e+14   2000\n",
       "4000004  3193581  315.241241  712.992493  2151.678467  1.149920e+14   2000"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tmp"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>px</th>\n",
       "      <th>py</th>\n",
       "      <th>pz</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>4000000</th>\n",
       "      <td>335.164948</td>\n",
       "      <td>742.002258</td>\n",
       "      <td>2128.719238</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4000001</th>\n",
       "      <td>320.469086</td>\n",
       "      <td>754.580933</td>\n",
       "      <td>2131.292969</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4000002</th>\n",
       "      <td>342.707184</td>\n",
       "      <td>787.741760</td>\n",
       "      <td>2101.839600</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4000003</th>\n",
       "      <td>326.419830</td>\n",
       "      <td>713.458313</td>\n",
       "      <td>2188.218262</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4000004</th>\n",
       "      <td>315.241241</td>\n",
       "      <td>712.992493</td>\n",
       "      <td>2151.678467</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                 px          py           pz\n",
       "4000000  335.164948  742.002258  2128.719238\n",
       "4000001  320.469086  754.580933  2131.292969\n",
       "4000002  342.707184  787.741760  2101.839600\n",
       "4000003  326.419830  713.458313  2188.218262\n",
       "4000004  315.241241  712.992493  2151.678467"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tmp[['px','py','pz']]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "4000000  px     335.164948\n",
       "         py     742.002258\n",
       "         pz    2128.719238\n",
       "4000001  px     320.469086\n",
       "         py     754.580933\n",
       "         pz    2131.292969\n",
       "4000002  px     342.707184\n",
       "         py     787.741760\n",
       "         pz    2101.839600\n",
       "4000003  px     326.419830\n",
       "         py     713.458313\n",
       "         pz    2188.218262\n",
       "4000004  px     315.241241\n",
       "         py     712.992493\n",
       "         pz    2151.678467\n",
       "dtype: float32"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tmp[['px','py','pz']].stack()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([ 335.16495,  742.00226, 2128.7192 ,  320.4691 ,  754.58093,\n",
       "       2131.293  ,  342.70718,  787.74176, 2101.8396 ,  326.41983,\n",
       "        713.4583 , 2188.2183 ,  315.24124,  712.9925 , 2151.6785 ],\n",
       "      dtype=float32)"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tmp[['px','py','pz']].stack().values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>haloid</th>\n",
       "      <th>halomass</th>\n",
       "      <th>px</th>\n",
       "      <th>py</th>\n",
       "      <th>pz</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>isect</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>2164471</td>\n",
       "      <td>7.075400e+14</td>\n",
       "      <td>113.276787</td>\n",
       "      <td>136.558777</td>\n",
       "      <td>158.994751</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>3540112</td>\n",
       "      <td>7.602910e+14</td>\n",
       "      <td>194.531143</td>\n",
       "      <td>100.442146</td>\n",
       "      <td>204.011581</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>1007418</td>\n",
       "      <td>5.546810e+14</td>\n",
       "      <td>199.341278</td>\n",
       "      <td>121.810860</td>\n",
       "      <td>357.886566</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>22408</td>\n",
       "      <td>1.148890e+15</td>\n",
       "      <td>121.965492</td>\n",
       "      <td>152.880112</td>\n",
       "      <td>489.232605</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>106870</td>\n",
       "      <td>9.824900e+14</td>\n",
       "      <td>111.221252</td>\n",
       "      <td>119.286346</td>\n",
       "      <td>520.265320</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        haloid      halomass          px          py          pz\n",
       "isect                                                           \n",
       "0      2164471  7.075400e+14  113.276787  136.558777  158.994751\n",
       "1      3540112  7.602910e+14  194.531143  100.442146  204.011581\n",
       "2      1007418  5.546810e+14  199.341278  121.810860  357.886566\n",
       "3        22408  1.148890e+15  121.965492  152.880112  489.232605\n",
       "4       106870  9.824900e+14  111.221252  119.286346  520.265320"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "grouped.nth(0).head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>haloid</th>\n",
       "      <th>px</th>\n",
       "      <th>py</th>\n",
       "      <th>pz</th>\n",
       "      <th>halomass</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>isect</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>2164471</td>\n",
       "      <td>113.276787</td>\n",
       "      <td>136.558777</td>\n",
       "      <td>158.994751</td>\n",
       "      <td>7.075400e+14</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>3540112</td>\n",
       "      <td>194.531143</td>\n",
       "      <td>100.442146</td>\n",
       "      <td>204.011581</td>\n",
       "      <td>7.602910e+14</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>1007418</td>\n",
       "      <td>199.341278</td>\n",
       "      <td>121.810860</td>\n",
       "      <td>357.886566</td>\n",
       "      <td>5.546810e+14</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>22408</td>\n",
       "      <td>121.965492</td>\n",
       "      <td>152.880112</td>\n",
       "      <td>489.232605</td>\n",
       "      <td>1.148890e+15</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>106870</td>\n",
       "      <td>111.221252</td>\n",
       "      <td>119.286346</td>\n",
       "      <td>520.265320</td>\n",
       "      <td>9.824900e+14</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        haloid          px          py          pz      halomass\n",
       "isect                                                           \n",
       "0      2164471  113.276787  136.558777  158.994751  7.075400e+14\n",
       "1      3540112  194.531143  100.442146  204.011581  7.602910e+14\n",
       "2      1007418  199.341278  121.810860  357.886566  5.546810e+14\n",
       "3        22408  121.965492  152.880112  489.232605  1.148890e+15\n",
       "4       106870  111.221252  119.286346  520.265320  9.824900e+14"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "grouped.first().head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>haloid</th>\n",
       "      <th>px</th>\n",
       "      <th>py</th>\n",
       "      <th>pz</th>\n",
       "      <th>halomass</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>isect</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>87076185</td>\n",
       "      <td>165.257690</td>\n",
       "      <td>123.588676</td>\n",
       "      <td>183.822357</td>\n",
       "      <td>2.083710e+12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>82113093</td>\n",
       "      <td>114.391266</td>\n",
       "      <td>164.690506</td>\n",
       "      <td>215.500046</td>\n",
       "      <td>2.128810e+12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>109595160</td>\n",
       "      <td>198.716934</td>\n",
       "      <td>134.817307</td>\n",
       "      <td>372.275604</td>\n",
       "      <td>2.525700e+12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>123007950</td>\n",
       "      <td>151.797165</td>\n",
       "      <td>105.821655</td>\n",
       "      <td>419.826538</td>\n",
       "      <td>2.092730e+12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>109357752</td>\n",
       "      <td>108.403801</td>\n",
       "      <td>116.367622</td>\n",
       "      <td>518.777649</td>\n",
       "      <td>2.219010e+12</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "          haloid          px          py          pz      halomass\n",
       "isect                                                             \n",
       "0       87076185  165.257690  123.588676  183.822357  2.083710e+12\n",
       "1       82113093  114.391266  164.690506  215.500046  2.128810e+12\n",
       "2      109595160  198.716934  134.817307  372.275604  2.525700e+12\n",
       "3      123007950  151.797165  105.821655  419.826538  2.092730e+12\n",
       "4      109357752  108.403801  116.367622  518.777649  2.219010e+12"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "grouped.last().head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>haloid</th>\n",
       "      <th>px</th>\n",
       "      <th>py</th>\n",
       "      <th>pz</th>\n",
       "      <th>halomass</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>isect</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>24784</td>\n",
       "      <td>100.000732</td>\n",
       "      <td>100.020172</td>\n",
       "      <td>100.005898</td>\n",
       "      <td>2.083710e+12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>16802</td>\n",
       "      <td>100.004387</td>\n",
       "      <td>100.051926</td>\n",
       "      <td>200.020035</td>\n",
       "      <td>2.128810e+12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>5906</td>\n",
       "      <td>100.097351</td>\n",
       "      <td>100.169243</td>\n",
       "      <td>300.020447</td>\n",
       "      <td>2.525700e+12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>22408</td>\n",
       "      <td>100.001022</td>\n",
       "      <td>100.023865</td>\n",
       "      <td>400.059113</td>\n",
       "      <td>2.092730e+12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>6075</td>\n",
       "      <td>100.022926</td>\n",
       "      <td>100.027275</td>\n",
       "      <td>500.028198</td>\n",
       "      <td>2.219010e+12</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       haloid          px          py          pz      halomass\n",
       "isect                                                          \n",
       "0       24784  100.000732  100.020172  100.005898  2.083710e+12\n",
       "1       16802  100.004387  100.051926  200.020035  2.128810e+12\n",
       "2        5906  100.097351  100.169243  300.020447  2.525700e+12\n",
       "3       22408  100.001022  100.023865  400.059113  2.092730e+12\n",
       "4        6075  100.022926  100.027275  500.028198  2.219010e+12"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "grouped.min().head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[0, 1, 2, 3, 4]"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "groupnamelist = grouped.groups.keys()\n",
    "groupnamelist[:5]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "> this is the `isect` values, the `groupby` keys; equivalently `hrdf.isect.unique()` "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([    0,     1,     2, ..., 26997, 26998, 26999])"
      ]
     },
     "execution_count": 29,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "hrdf['isect'].unique()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Extract top 400 halos in mass with various features. "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 20000 training set, using `Spark` for parallel UDF computations\n",
    "\n",
    "> The size of `hrdf` is small enought to fit in a single machine. However, the `apply`ing takes too much time. Hence, I call `Spark` for help for parallel computation. `PandasUDF` will be applied. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "from pyspark import SparkContext   \n",
    "from pyspark.sql import SQLContext\n",
    "\n",
    "#sc = SparkContext(master='local[3]', appName='calgraph')\n",
    "sqlsc = SQLContext(sc)\n",
    "#sc.setCheckpointDir(\"./checkpoints\")\n",
    "#sc.setCheckpointDir(\"hdfs://localhost:8020/myhdfs/spark/checkpoints\")\n",
    "sc.setCheckpointDir(\"hdfs://master:54310/tmp/spark/checkpoints\")\n",
    "\n",
    "import pyspark.sql.functions as F\n",
    "import pyspark.sql.types as T\n",
    "from pyspark import Row\n",
    "from pyspark.sql.window import Window as W\n",
    "\n",
    "spark.conf.set(\"spark.sql.execution.arrow.enabled\", \"true\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "> Create a spark df from the pandas df "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [],
   "source": [
    "# get a spark dataframe from pandas dataframe\n",
    "shrdf = sqlsc.createDataFrame(hrdf)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+-------+---------+---------+---------+-------------+-----+\n",
      "| haloid|       px|       py|       pz|     halomass|isect|\n",
      "+-------+---------+---------+---------+-------------+-----+\n",
      "|2164471|113.27679|136.55878|158.99475| 7.0754003E14|    0|\n",
      "| 341850| 148.6161|193.99586|172.71184| 6.1874399E14|    0|\n",
      "|2654219|156.18562|115.29642|168.60387| 5.6340401E14|    0|\n",
      "|1339587|188.42734|189.25693|168.73097|3.79027005E14|    0|\n",
      "|  24784|145.33125|188.97966|175.40134|3.55501993E14|    0|\n",
      "+-------+---------+---------+---------+-------------+-----+\n",
      "only showing top 5 rows\n",
      "\n"
     ]
    }
   ],
   "source": [
    "shrdf.show(5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "root\n",
      " |-- haloid: integer (nullable = true)\n",
      " |-- px: float (nullable = true)\n",
      " |-- py: float (nullable = true)\n",
      " |-- pz: float (nullable = true)\n",
      " |-- halomass: float (nullable = true)\n",
      " |-- isect: long (nullable = true)\n",
      "\n"
     ]
    }
   ],
   "source": [
    "shrdf.printSchema()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>haloid</th>\n",
       "      <th>px</th>\n",
       "      <th>py</th>\n",
       "      <th>pz</th>\n",
       "      <th>halomass</th>\n",
       "      <th>isect</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>2164471</td>\n",
       "      <td>113.276787</td>\n",
       "      <td>136.558777</td>\n",
       "      <td>158.994751</td>\n",
       "      <td>7.075400e+14</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>341850</td>\n",
       "      <td>148.616104</td>\n",
       "      <td>193.995865</td>\n",
       "      <td>172.711838</td>\n",
       "      <td>6.187440e+14</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2654219</td>\n",
       "      <td>156.185623</td>\n",
       "      <td>115.296417</td>\n",
       "      <td>168.603867</td>\n",
       "      <td>5.634040e+14</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>1339587</td>\n",
       "      <td>188.427338</td>\n",
       "      <td>189.256927</td>\n",
       "      <td>168.730972</td>\n",
       "      <td>3.790270e+14</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>24784</td>\n",
       "      <td>145.331253</td>\n",
       "      <td>188.979660</td>\n",
       "      <td>175.401337</td>\n",
       "      <td>3.555020e+14</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    haloid          px          py          pz      halomass  isect\n",
       "0  2164471  113.276787  136.558777  158.994751  7.075400e+14      0\n",
       "1   341850  148.616104  193.995865  172.711838  6.187440e+14      0\n",
       "2  2654219  156.185623  115.296417  168.603867  5.634040e+14      0\n",
       "3  1339587  188.427338  189.256927  168.730972  3.790270e+14      0\n",
       "4    24784  145.331253  188.979660  175.401337  3.555020e+14      0"
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "hrdf.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [],
   "source": [
    "numhalo = 400 # 400 points for each box "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [],
   "source": [
    "udfschema = T.StructType([\\\n",
    "                          T.StructField('haloid', T.IntegerType(), False),\\\n",
    "                          T.StructField('px', T.FloatType(), False),\\\n",
    "                          T.StructField('py', T.FloatType(), False),\\\n",
    "                          T.StructField('pz', T.FloatType(), False),\\\n",
    "                          T.StructField('halomass', T.FloatType(), False),\\\n",
    "                          T.StructField('isect', T.LongType(), False),\\\n",
    "                          T.StructField('srank', T.FloatType(), False)\\\n",
    "                         ])                        "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "> Define a pandas_udf `extractFeature`"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [],
   "source": [
    "from pyspark.sql.functions import pandas_udf, PandasUDFType"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Input dataframe schema\n",
    "@pandas_udf(udfschema, PandasUDFType.GROUPED_MAP)\n",
    "def extractFeature(eachgroupdf):\n",
    "    tmpdf = eachgroupdf[:400] # 400 halos; maybe, we may find another way to pass this argurment `numhalo=400` later\n",
    "    \n",
    "    ## normalize the coordinates in 1x1x1 cube \n",
    "    tmpdf.loc[:,['px','py','pz']] = tmpdf.loc[:,['px','py','pz']].abs() # make some negative coordinate glitches positive forcefully. \n",
    "    tmpdf.px = 0.01*(tmpdf.px - np.floor(tmpdf.px.min())) #normalize 100Mpc to 1, unity\n",
    "    tmpdf.py = 0.01*(tmpdf.py - np.floor(tmpdf.py.min()))\n",
    "    tmpdf.pz = 0.01*(tmpdf.pz - np.floor(tmpdf.pz.min()))\n",
    "    \n",
    "    tmpdf['srank'] = tmpdf\\\n",
    "    .apply(lambda row: np.sqrt((row['px']-0.5)**2 +(row['px']-0.5)**2 +(row['px']-0.5)**2), axis=1)\n",
    "    tmpdf = tmpdf.sort_values(by='srank').reset_index(drop=True)\n",
    "    \n",
    "    return tmpdf"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "> Test `extractFeature`"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [],
   "source": [
    "sample = shrdf.filter(shrdf.isect == 0).toPandas()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>haloid</th>\n",
       "      <th>px</th>\n",
       "      <th>py</th>\n",
       "      <th>pz</th>\n",
       "      <th>halomass</th>\n",
       "      <th>isect</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>2164471</td>\n",
       "      <td>113.276787</td>\n",
       "      <td>136.558777</td>\n",
       "      <td>158.994751</td>\n",
       "      <td>7.075400e+14</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>341850</td>\n",
       "      <td>148.616104</td>\n",
       "      <td>193.995865</td>\n",
       "      <td>172.711838</td>\n",
       "      <td>6.187440e+14</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2654219</td>\n",
       "      <td>156.185623</td>\n",
       "      <td>115.296417</td>\n",
       "      <td>168.603867</td>\n",
       "      <td>5.634040e+14</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>1339587</td>\n",
       "      <td>188.427338</td>\n",
       "      <td>189.256927</td>\n",
       "      <td>168.730972</td>\n",
       "      <td>3.790270e+14</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>24784</td>\n",
       "      <td>145.331253</td>\n",
       "      <td>188.979660</td>\n",
       "      <td>175.401337</td>\n",
       "      <td>3.555020e+14</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    haloid          px          py          pz      halomass  isect\n",
       "0  2164471  113.276787  136.558777  158.994751  7.075400e+14      0\n",
       "1   341850  148.616104  193.995865  172.711838  6.187440e+14      0\n",
       "2  2654219  156.185623  115.296417  168.603867  5.634040e+14      0\n",
       "3  1339587  188.427338  189.256927  168.730972  3.790270e+14      0\n",
       "4    24784  145.331253  188.979660  175.401337  3.555020e+14      0"
      ]
     },
     "execution_count": 40,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sample.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "RangeIndex: 2000 entries, 0 to 1999\n",
      "Data columns (total 6 columns):\n",
      "haloid      2000 non-null int32\n",
      "px          2000 non-null float32\n",
      "py          2000 non-null float32\n",
      "pz          2000 non-null float32\n",
      "halomass    2000 non-null float32\n",
      "isect       2000 non-null int64\n",
      "dtypes: float32(4), int32(1), int64(1)\n",
      "memory usage: 54.8 KB\n"
     ]
    }
   ],
   "source": [
    "sample.info(verbose=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/shong/anaconda2/lib/python2.7/site-packages/pandas/core/indexing.py:543: SettingWithCopyWarning: \n",
      "A value is trying to be set on a copy of a slice from a DataFrame.\n",
      "Try using .loc[row_indexer,col_indexer] = value instead\n",
      "\n",
      "See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/indexing.html#indexing-view-versus-copy\n",
      "  self.obj[item] = s\n",
      "/home/shong/anaconda2/lib/python2.7/site-packages/pandas/core/generic.py:4405: SettingWithCopyWarning: \n",
      "A value is trying to be set on a copy of a slice from a DataFrame.\n",
      "Try using .loc[row_indexer,col_indexer] = value instead\n",
      "\n",
      "See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/indexing.html#indexing-view-versus-copy\n",
      "  self[name] = value\n",
      "/home/shong/anaconda2/lib/python2.7/site-packages/ipykernel_launcher.py:12: SettingWithCopyWarning: \n",
      "A value is trying to be set on a copy of a slice from a DataFrame.\n",
      "Try using .loc[row_indexer,col_indexer] = value instead\n",
      "\n",
      "See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/indexing.html#indexing-view-versus-copy\n",
      "  if sys.path[0] == '':\n"
     ]
    }
   ],
   "source": [
    "samplefeature = extractFeature.func(sample)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "RangeIndex: 400 entries, 0 to 399\n",
      "Data columns (total 7 columns):\n",
      "haloid      400 non-null int32\n",
      "px          400 non-null float32\n",
      "py          400 non-null float32\n",
      "pz          400 non-null float32\n",
      "halomass    400 non-null float32\n",
      "isect       400 non-null int64\n",
      "srank       400 non-null float64\n",
      "dtypes: float32(4), float64(1), int32(1), int64(1)\n",
      "memory usage: 14.1 KB\n"
     ]
    }
   ],
   "source": [
    "samplefeature.info(verbose=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/shong/anaconda2/lib/python2.7/site-packages/ipykernel_launcher.py:12: SettingWithCopyWarning: \n",
      "A value is trying to be set on a copy of a slice from a DataFrame.\n",
      "Try using .loc[row_indexer,col_indexer] = value instead\n",
      "\n",
      "See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/indexing.html#indexing-view-versus-copy\n",
      "  if sys.path[0] == '':\n"
     ]
    }
   ],
   "source": [
    "spksample = sqlsc.createDataFrame(extractFeature.func(sample))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "root\n",
      " |-- haloid: integer (nullable = true)\n",
      " |-- px: float (nullable = true)\n",
      " |-- py: float (nullable = true)\n",
      " |-- pz: float (nullable = true)\n",
      " |-- halomass: float (nullable = true)\n",
      " |-- isect: long (nullable = true)\n",
      " |-- srank: double (nullable = true)\n",
      "\n"
     ]
    }
   ],
   "source": [
    "spksample.printSchema()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "> Apply `extractFeature` to all spark df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [],
   "source": [
    "featuredf = shrdf.groupby('isect').apply(extractFeature)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+--------+----------+----------+------------+-------------+-----+------------+\n",
      "|  haloid|        px|        py|          pz|     halomass|isect|       srank|\n",
      "+--------+----------+----------+------------+-------------+-----+------------+\n",
      "| 3057075|  0.499467|0.78901166|   0.8512012| 6.8825498E12|   26|9.2320825E-4|\n",
      "| 2178090|0.49938232|0.78155667|0.0033618163| 6.2502201E13|   26|0.0010698582|\n",
      "| 6851769|0.49920136|  0.252165|   0.5463476|1.25473003E13|   26|0.0013832896|\n",
      "|27415387|0.49891815|0.18716209|  0.87591064|1.45317995E13|   26|0.0018738263|\n",
      "|14870352| 0.5019577| 0.6532689|  0.02826416|1.98447999E13|   26| 0.003390861|\n",
      "| 9951636| 0.5025119|0.93468064|  0.61061764| 7.7304598E12|   26|0.0043507707|\n",
      "| 7119205| 0.5063745|0.31730726|  0.89315915| 9.9557904E13|   26|  0.01104092|\n",
      "|10074734|  0.507727|0.69688416|  0.41511962|   1.24932E13|   26| 0.013383604|\n",
      "|15081707| 0.5134178|0.04084877|   0.1903711|1.73371996E13|   26| 0.023240278|\n",
      "|35314192| 0.5134691|0.17705505|   0.6345068|2.59335997E13|   26| 0.023329165|\n",
      "+--------+----------+----------+------------+-------------+-----+------------+\n",
      "only showing top 10 rows\n",
      "\n",
      "CPU times: user 6.54 ms, sys: 3.41 ms, total: 9.94 ms\n",
      "Wall time: 39.8 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "featuredf.show(10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "root\n",
      " |-- haloid: integer (nullable = false)\n",
      " |-- px: float (nullable = false)\n",
      " |-- py: float (nullable = false)\n",
      " |-- pz: float (nullable = false)\n",
      " |-- halomass: float (nullable = false)\n",
      " |-- isect: long (nullable = false)\n",
      " |-- srank: float (nullable = false)\n",
      "\n"
     ]
    }
   ],
   "source": [
    "featuredf.printSchema()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Now, execute the pandasUDF... "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 250 ms, sys: 395 ms, total: 645 ms\n",
      "Wall time: 4min 52s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "featurepdf = featuredf.toPandas()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "> Okey, now it only takes 5 min. It took several hours when using a naive single-machine python loop. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "10800000"
      ]
     },
     "execution_count": 50,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(featurepdf.index)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "count    1.080000e+07\n",
       "mean     1.349950e+04\n",
       "std      7.794229e+03\n",
       "min      0.000000e+00\n",
       "25%      6.749750e+03\n",
       "50%      1.349950e+04\n",
       "75%      2.024925e+04\n",
       "max      2.699900e+04\n",
       "Name: isect, dtype: float64"
      ]
     },
     "execution_count": 51,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "featurepdf['isect'].describe()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>haloid</th>\n",
       "      <th>px</th>\n",
       "      <th>py</th>\n",
       "      <th>pz</th>\n",
       "      <th>halomass</th>\n",
       "      <th>isect</th>\n",
       "      <th>srank</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>3057075</td>\n",
       "      <td>0.499467</td>\n",
       "      <td>0.789012</td>\n",
       "      <td>0.851201</td>\n",
       "      <td>6.882550e+12</td>\n",
       "      <td>26</td>\n",
       "      <td>0.000923</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2178090</td>\n",
       "      <td>0.499382</td>\n",
       "      <td>0.781557</td>\n",
       "      <td>0.003362</td>\n",
       "      <td>6.250220e+13</td>\n",
       "      <td>26</td>\n",
       "      <td>0.001070</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>6851769</td>\n",
       "      <td>0.499201</td>\n",
       "      <td>0.252165</td>\n",
       "      <td>0.546348</td>\n",
       "      <td>1.254730e+13</td>\n",
       "      <td>26</td>\n",
       "      <td>0.001383</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>27415387</td>\n",
       "      <td>0.498918</td>\n",
       "      <td>0.187162</td>\n",
       "      <td>0.875911</td>\n",
       "      <td>1.453180e+13</td>\n",
       "      <td>26</td>\n",
       "      <td>0.001874</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>14870352</td>\n",
       "      <td>0.501958</td>\n",
       "      <td>0.653269</td>\n",
       "      <td>0.028264</td>\n",
       "      <td>1.984480e+13</td>\n",
       "      <td>26</td>\n",
       "      <td>0.003391</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "     haloid        px        py        pz      halomass  isect     srank\n",
       "0   3057075  0.499467  0.789012  0.851201  6.882550e+12     26  0.000923\n",
       "1   2178090  0.499382  0.781557  0.003362  6.250220e+13     26  0.001070\n",
       "2   6851769  0.499201  0.252165  0.546348  1.254730e+13     26  0.001383\n",
       "3  27415387  0.498918  0.187162  0.875911  1.453180e+13     26  0.001874\n",
       "4  14870352  0.501958  0.653269  0.028264  1.984480e+13     26  0.003391"
      ]
     },
     "execution_count": 52,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "featurepdf.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [],
   "source": [
    "groupfeaturepdf = featurepdf.groupby('isect')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[0, 1, 2, 3, 4]"
      ]
     },
     "execution_count": 54,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "groupfeaturepdf.groups.keys()[:5]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "RangeIndex: 10800000 entries, 0 to 10799999\n",
      "Data columns (total 7 columns):\n",
      "haloid      int32\n",
      "px          float32\n",
      "py          float32\n",
      "pz          float32\n",
      "halomass    float32\n",
      "isect       int64\n",
      "srank       float32\n",
      "dtypes: float32(5), int32(1), int64(1)\n",
      "memory usage: 329.6 MB\n"
     ]
    }
   ],
   "source": [
    "featurepdf.info()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 7000 test set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "hrtest =[]\n",
    "\n",
    "for isample in range(20000,27000):\n",
    "#for isample in range(20000,20005):\n",
    "    tmpdf = hrdf[hrdf.isect == isample][:numhalo]\n",
    "    ## normalize the coordinates in 1x1x1 cube \n",
    "    tmpdf = tmpdf.abs() # some coordinate glitches : which have negative values. \n",
    "    tmpdf.px = 0.01*(tmpdf.px - np.floor(tmpdf.px.min()))\n",
    "    tmpdf.py = 0.01*(tmpdf.py - np.floor(tmpdf.py.min()))\n",
    "    tmpdf.pz = 0.01*(tmpdf.pz - np.floor(tmpdf.pz.min()))\n",
    "    \n",
    "    tmpdf['srank'] = tmpdf\\\n",
    "    .apply(lambda row: np.sqrt((row['px']-0.5)**2 +(row['px']-0.5)**2 +(row['px']-0.5)**2), axis=1)\n",
    "    tmpdf = tmpdf.sort_values(by='srank').reset_index(drop=True)\n",
    "    tmp = tmpdf[['px','py','pz']].stack()\n",
    "    hrtest.append(tmp.values)\n",
    "\n",
    "hrtest = np.float32(hrtest)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(7000, 1200)"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "hrtest.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Generate Random Sets \n",
    "\n",
    "> 20000 train 7000 test samples"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "nTrainSamples = 20000\n",
    "nTestSamples = 7000"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "def genRandomCube(nsize):\n",
    "    re = []\n",
    "    for i in range(nsize): \n",
    "        tmpPoint = np.random.random(3)\n",
    "        re.append(tmpPoint)\n",
    "    return np.float32(re)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "rdtrain = []\n",
    "isample = 0\n",
    "\n",
    "while isample < nTrainSamples:\n",
    "    tmpdf = pd.DataFrame(genRandomCube(numhalo),columns=['px','py','pz'])\n",
    "    \n",
    "    tmpdf['srank'] = tmpdf\\\n",
    "    .apply(lambda row: np.sqrt((row['px']-0.5)**2 +(row['px']-0.5)**2 +(row['px']-0.5)**2), axis=1)\n",
    "    tmpdf = tmpdf.sort_values(by='srank').reset_index(drop=True)\n",
    "    tmp = tmpdf[['px','py','pz']].stack()\n",
    "    rdtrain.append(tmp.values)\n",
    "    isample = isample+1\n",
    "\n",
    "rdtrain = np.float32(rdtrain)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(20000, 1200)"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "rdtrain.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "rdtest = []\n",
    "isample = 0\n",
    "\n",
    "while isample < nTestSamples:\n",
    "    tmpdf = pd.DataFrame(genRandomCube(numhalo),columns=['px','py','pz'])\n",
    "    tmpdf['srank'] = tmpdf\\\n",
    "    .apply(lambda row: np.sqrt((row['px']-0.5)**2 +(row['px']-0.5)**2 +(row['px']-0.5)**2), axis=1)\n",
    "    tmpdf = tmpdf.sort_values(by='srank').reset_index(drop=True)\n",
    "    tmp = tmpdf[['px','py','pz']].stack()\n",
    "    rdtest.append(tmp.values)\n",
    "    isample = isample+1\n",
    "\n",
    "rdtest = np.float32(rdtest)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(7000, 1200)"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "rdtest.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "> Plot samples "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([0.49855694, 0.821905  , 0.06488563, 0.49749225, 0.17175856,\n",
       "       0.49279207, 0.50519747, 0.92783904, 0.6619252 ], dtype=float32)"
      ]
     },
     "execution_count": 29,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "rdtrain[0,:9]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "ix = np.arange(numhalo)*3\n",
    "iy = np.arange(numhalo)*3 + 1\n",
    "iz = np.arange(numhalo)*3 + 2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "from mpl_toolkits.mplot3d import Axes3D\n",
    "\n",
    "# plot settings\n",
    "plt.rc('font', family='serif') \n",
    "plt.rc('font', serif='Times New Roman') \n",
    "plt.rcParams.update({'font.size': 14})\n",
    "plt.rcParams['mathtext.fontset'] = 'stix'\n",
    "\n",
    "fig = plt.figure(figsize=(8,4))\n",
    "\n",
    "ax = fig.add_subplot(121, projection='3d')\n",
    "ax.set_title(r'Halos: $N = 400$')\n",
    "ax.set_xlim(0, 1)\n",
    "ax.set_ylim(0, 1)\n",
    "ax.set_zlim(0, 1)\n",
    "\n",
    "ax.scatter(hrtrain[0,ix],hrtrain[0,iy],hrtrain[0,iz])\n",
    "ax.set_xlabel(r'$x$',fontsize=22)\n",
    "ax.set_ylabel(r'$y$',fontsize=22)\n",
    "ax.set_zlabel(r'$z$',fontsize=22)\n",
    "\n",
    "ax = fig.add_subplot(122, projection='3d')\n",
    "ax.set_title(r'Random: $N = 400$')\n",
    "ax.set_xlim(0, 1)\n",
    "ax.set_ylim(0, 1)\n",
    "ax.set_zlim(0, 1)\n",
    "\n",
    "ax.scatter(rdtrain[1,ix],rdtrain[1,iy],rdtrain[1,iz])\n",
    "ax.set_xlabel(r'$x$',fontsize=22)\n",
    "ax.set_ylabel(r'$y$',fontsize=22)\n",
    "ax.set_zlabel(r'$z$',fontsize=22)\n",
    "\n",
    "\n",
    "\n",
    "\n",
    "plt.tight_layout(pad=0.4, w_pad=0.5, h_pad=1.0)\n",
    "plt.savefig('sample400.png')\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Concatenate the two kinds of samples to a single train/test set"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Train Set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train = np.concatenate((hrtrain,rdtrain),axis=0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(40000, 1200)"
      ]
     },
     "execution_count": 33,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_train.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "> Classification Code : HR = 0, Random = 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_train = np.concatenate((np.full(nTrainSamples,0),np.full(nTrainSamples,1)),axis=0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(40000,)"
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_train.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "> Randomize the train set "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [],
   "source": [
    "lentrain = len(y_train)\n",
    "irand = np.arange(lentrain)\n",
    "np.random.seed(7)\n",
    "np.random.shuffle(irand)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_trainRand = []\n",
    "y_trainRand = []\n",
    "\n",
    "for ir in irand:\n",
    "    X_trainRand.append(X_train[ir])\n",
    "    y_trainRand.append(y_train[ir])\n",
    "\n",
    "X_trainRand = np.float32(X_trainRand)\n",
    "y_trainRand = np.int32(y_trainRand)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(40000, 1200)"
      ]
     },
     "execution_count": 38,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_trainRand.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(40000,)"
      ]
     },
     "execution_count": 39,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_trainRand.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pickle\n",
    "\n",
    "with open('trainRand400.pickle','wb') as f:\n",
    "    pickle.dump([X_trainRand, y_trainRand],f)\n",
    "f.close() #keep dumping the current results to overwrite the pickle. "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Test Set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_test = np.concatenate((hrtest,rdtest),axis=0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(14000, 1200)"
      ]
     },
     "execution_count": 42,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_test.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_test = np.concatenate((np.full(nTestSamples,0),np.full(nTestSamples,1)),axis=0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(14000,)"
      ]
     },
     "execution_count": 44,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_test.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [],
   "source": [
    "lentest = len(y_test)\n",
    "irand = np.arange(lentest)\n",
    "np.random.seed(9)\n",
    "np.random.shuffle(irand)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_testRand = []\n",
    "y_testRand = []\n",
    "\n",
    "for ir in irand:\n",
    "    X_testRand.append(X_test[ir])\n",
    "    y_testRand.append(y_test[ir])\n",
    "\n",
    "X_testRand = np.float32(X_testRand)\n",
    "y_testRand = np.int32(y_testRand)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(14000, 1200)"
      ]
     },
     "execution_count": 47,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_testRand.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(14000,)"
      ]
     },
     "execution_count": 48,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_testRand.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [],
   "source": [
    "with open('testRand400.pickle','wb') as f:\n",
    "    pickle.dump([X_testRand, y_testRand],f)\n",
    "f.close() #keep dumping the current results to overwrite the pickle. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "u'/Users/shong/work/deeplearning/cosmo-pattern'"
      ]
     },
     "execution_count": 50,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pwd"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 2",
   "language": "python",
   "name": "python2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.15"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}