Quora_Answer.ipynb

tf-prob-example

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[0.16666666666666666,\n",
       " 0.16666666666666666,\n",
       " 0.16666666666666666,\n",
       " 0.16666666666666666,\n",
       " 0.16666666666666666,\n",
       " 0.16666666666666666]"
      ]
     },
     "execution_count": 1,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#import important libs\n",
    "import tensorflow as tf\n",
    "import tensorflow_probability as tfp\n",
    "import numpy as np\n",
    "# Set the mode to eager execution\n",
    "tf.enable_eager_execution()\n",
    "\n",
    "# We have a fair die, so the probability of each side is equally likely\n",
    "probs = [1/6.] * 6\n",
    "probs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "# This is where we are going to store the final result of each experiment\n",
    "var_arr = []"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "# We do the same experiment 100 times\n",
    "for k in range(0, 100):\n",
    "    # Each time we throw our fair die 1000 times and record how many times each side is shown\n",
    "    dices = tfp.distributions.Multinomial(total_count=1000, probs=probs)\n",
    "    n = dices.sample()\n",
    "    # Then conver the tf.tensor to a numpy array\n",
    "    l = n.numpy()\n",
    "    L = []\n",
    "    # For each side of the die\n",
    "    for i, nums in enumerate(l):\n",
    "        # For how many times that side has turned up\n",
    "        for _ in range(int(nums)):\n",
    "            # if it is an odd number fase of the die (1, 3, 5) then toss the fair coin 3 times else 2 times\n",
    "            mul_by = 3 if (i + 1) % 2 != 0 else 2\n",
    "            # It is a fair coin thus each side is equally probable\n",
    "            tosses = tfp.distributions.Bernoulli(probs=0.5)\n",
    "            coin_flip_data = tosses.sample(mul_by)\n",
    "            l2 = coin_flip_data.numpy()\n",
    "            # How many times head turned up (assuming HEAD = 1)\n",
    "            num_heads = np.sum(l2)\n",
    "            L += [num_heads]\n",
    "    # Calculate the variance and store it\n",
    "    var_arr.append(np.var(L))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.68355216"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.mean(var_arr)"
   ]
  }
 ],
 "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
}