Probabilistic Machine Learning/Latent Dirichlet Allocation/Latent Dirichlet Allocation.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import scipy.io as sio\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "from scipy.stats import norm\n",
    "from scipy.sparse import csc_matrix\n",
    "import math\n",
    "mat_contents = sio.loadmat('kos_doc_data.mat')\n",
    "mat_contents2= sio.loadmat('store_b.mat')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [],
   "source": [
    "def sampDiscrete(b):\n",
    "    r = np.sum(b)*np.random.uniform(0,1,1)\n",
    "    a = b[0]; i = 0\n",
    "    while a < r:\n",
    "        i+=1\n",
    "        if i == 20-1:\n",
    "            break\n",
    "        a+=b[i]\n",
    "    return i"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Min ID of documents in A: 1\n",
      "Max ID of documents in A: 2000\n",
      "No. of distinct words from A: 6906\n",
      "No. of distinct word occurences in As doc: 29\n",
      "Min ID of documents in B: 2001\n",
      "Max ID of documents in B: 3430\n",
      "No. of distinct words from B: 6906\n",
      "No. of distinct word occurences in Bs doc: 43\n",
      "No. of distinct words frBom V: 6906\n"
     ]
    }
   ],
   "source": [
    "A = mat_contents['A'] # train\n",
    "B = mat_contents['B'] # test\n",
    "V = mat_contents['V'] # vocabulary list\n",
    "\n",
    "print('Min ID of documents in A: '+str(A[:,0].min()))\n",
    "print('Max ID of documents in A: '+str(A[:,0].max()))\n",
    "print('No. of distinct words from A: '+str(A[:,1].max()))\n",
    "print('No. of distinct word occurences in As doc: '+str(A[:,2].max()))\n",
    "\n",
    "print('Min ID of documents in B: '+str(B[:,0].min()))\n",
    "print('Max ID of documents in B: '+str(B[:,0].max()))\n",
    "print('No. of distinct words from B: '+str(B[:,1].max()))\n",
    "print('No. of distinct word occurences in Bs doc: '+str(B[:,2].max()))\n",
    "\n",
    "print('No. of distinct words frBom V: '+str(len(V)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(1, 6906)\n"
     ]
    }
   ],
   "source": [
    "store_b = mat_contents2['b']\n",
    "print(store_b.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "W = A[:,1].max(); D = A[:,0].max(); K = 20\n",
    "alpha = .1; gamma = .1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [],
   "source": [
    "swd = csc_matrix((A[:,2],(A[:,1]-1,A[:,0]-1)))\n",
    "Swd = csc_matrix((B[:,2],(B[:,1]-1,B[:,0]-1)))\n",
    "skd = np.zeros((K,D))\n",
    "swk = np.zeros((W,K))\n",
    "s = []\n",
    "for d in range(D):\n",
    "    z = np.zeros((W,K))\n",
    "    for w in A[A[:,0]==d+1,1]:\n",
    "        c = swd[w-1,d]\n",
    "        for i in range(c):\n",
    "            k = np.ceil(K*np.random.uniform(0,1,1)).astype(int)\n",
    "            z[w-1,k-1] += 1\n",
    "    skd[:,d] = np.sum(z,axis=0)   \n",
    "    swk += z\n",
    "    s.append(csc_matrix(z))\n",
    "    \n",
    "sk = np.sum(skd,axis=1)\n",
    "topic_mixpro = []; entropy = np.zeros((20,20))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {},
   "outputs": [
    {
     "ename": "KeyboardInterrupt",
     "evalue": "",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mKeyboardInterrupt\u001b[0m                         Traceback (most recent call last)",
      "\u001b[1;32m<ipython-input-58-a47b65ec9bcc>\u001b[0m in \u001b[0;36m<module>\u001b[1;34m\u001b[0m\n\u001b[0;32m      9\u001b[0m                     \u001b[0msk\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0mk\u001b[0m\u001b[1;33m-\u001b[0m\u001b[1;36m1\u001b[0m\u001b[1;33m]\u001b[0m \u001b[1;33m-=\u001b[0m \u001b[1;36m1\u001b[0m\u001b[1;33m;\u001b[0m \u001b[0mskd\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0mk\u001b[0m\u001b[1;33m-\u001b[0m\u001b[1;36m1\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0md\u001b[0m\u001b[1;33m]\u001b[0m \u001b[1;33m-=\u001b[0m \u001b[1;36m1\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     10\u001b[0m                     \u001b[0mb\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0malpha\u001b[0m\u001b[1;33m+\u001b[0m\u001b[0mskd\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0md\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m*\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mgamma\u001b[0m\u001b[1;33m+\u001b[0m\u001b[0mswk\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0mw\u001b[0m\u001b[1;33m-\u001b[0m\u001b[1;36m1\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m/\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mW\u001b[0m\u001b[1;33m*\u001b[0m\u001b[0mgamma\u001b[0m\u001b[1;33m+\u001b[0m\u001b[0msk\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m---> 11\u001b[1;33m                     \u001b[0mkk\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0msampDiscrete\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mb\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m     12\u001b[0m                     \u001b[0mz\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0mw\u001b[0m\u001b[1;33m-\u001b[0m\u001b[1;36m1\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0mkk\u001b[0m\u001b[1;33m]\u001b[0m \u001b[1;33m+=\u001b[0m \u001b[1;36m1\u001b[0m\u001b[1;33m;\u001b[0m \u001b[0mskd\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0mkk\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0md\u001b[0m\u001b[1;33m]\u001b[0m \u001b[1;33m+=\u001b[0m \u001b[1;36m1\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     13\u001b[0m                     \u001b[0mswk\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0mw\u001b[0m\u001b[1;33m-\u001b[0m\u001b[1;36m1\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0mkk\u001b[0m\u001b[1;33m]\u001b[0m \u001b[1;33m+=\u001b[0m \u001b[1;36m1\u001b[0m\u001b[1;33m;\u001b[0m \u001b[0msk\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0mkk\u001b[0m\u001b[1;33m]\u001b[0m \u001b[1;33m+=\u001b[0m \u001b[1;36m1\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m<ipython-input-35-c1ca34595d00>\u001b[0m in \u001b[0;36msampDiscrete\u001b[1;34m(b)\u001b[0m\n\u001b[0;32m      1\u001b[0m \u001b[1;32mdef\u001b[0m \u001b[0msampDiscrete\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mb\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m----> 2\u001b[1;33m     \u001b[0mr\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mnp\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0msum\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mb\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m*\u001b[0m\u001b[0mnp\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mrandom\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0muniform\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;36m0\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;36m1\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;36m1\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m      3\u001b[0m     \u001b[0ma\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mb\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;36m0\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m;\u001b[0m \u001b[0mi\u001b[0m \u001b[1;33m=\u001b[0m \u001b[1;36m0\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m      4\u001b[0m     \u001b[1;32mwhile\u001b[0m \u001b[0ma\u001b[0m \u001b[1;33m<\u001b[0m \u001b[0mr\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m      5\u001b[0m         \u001b[0mi\u001b[0m\u001b[1;33m+=\u001b[0m\u001b[1;36m1\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;31mKeyboardInterrupt\u001b[0m: "
     ]
    }
   ],
   "source": [
    "for itr in range(100):\n",
    "    for d in range(D):\n",
    "        z = s[d].toarray()\n",
    "        for w in A[A[:,0]==d+1,1]:\n",
    "            a = z[w-1,:]; ka = np.where(a)\n",
    "            for k in ka[0][np.random.permutation(len(ka))]:\n",
    "                for i in range(a[k].astype(int)):\n",
    "                    z[w-1,k-1] -= 1; swk[w-1,k-1] -= 1\n",
    "                    sk[k-1] -= 1; skd[k-1,d] -= 1\n",
    "                    b=(alpha+skd[:,d])*(gamma+swk[w-1,:])/(W*gamma+sk)\n",
    "                    kk = sampDiscrete(b)\n",
    "                    z[w-1,kk] += 1; skd[kk,d] += 1\n",
    "                    swk[w-1,kk] += 1; sk[kk] += 1   \n",
    "        s[d] = csc_matrix(z)\n",
    "    topic_mixpro.append(sk/np.sum(sk))\n",
    "    for j in range(20):\n",
    "        topic_s = swk[:,j]\n",
    "        en = 0\n",
    "        for wo in range(6906):\n",
    "            if topic_s[wo] == 0:\n",
    "                en = en\n",
    "            else:\n",
    "                en -= store_b[0,wo]*np.log(store_b[0,wo])\n",
    "        entropy[j,itr] = en\n",
    "\n",
    "#     break"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1440x1080 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(20,15))\n",
    "for i in range(20):\n",
    "    plt.plot(np.arange(20),entropy[i,:])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "14"
      ]
     },
     "execution_count": 60,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "itr"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "lp = 0; nd = 0;\n",
    "for d in np.unique(B[:,0]):\n",
    "    z = np.zeros((W,K))\n",
    "    for w in B[B[:,0]==d+1,1]:\n",
    "        for i in range(Swd[w-1,d]):\n",
    "            k = np.ceil(K*np.random.uniform(0,1,1)).astype(int)\n",
    "            z[w-1,k-1] += 1\n",
    "    Skd = np.sum(z,axis=0)\n",
    "    Sk = sk+Skd\n",
    "    for itr in range(100):\n",
    "        for w in B[B[:,0]==d+1,1]:\n",
    "            a = z[w-1,:]; ka = np.where(a)\n",
    "            for k in ka[0][np.random.permutation(len(ka))]:\n",
    "                for i in range(a[k].astype(int)):\n",
    "                    z[w-1,k-1] -= 1; Skd[k-1] -= 1\n",
    "                    b=(alpha+Skd)*(gamma+swk[w-1,:])/(W*gamma+sk)\n",
    "                    kk = sampDiscrete(b)\n",
    "                    z[w-1,kk] += 1; Skd[kk] += 1\n",
    "    b=(alpha+Skd)/np.sum(alpha+Skd)*((gamma+swk)/(W*gamma+sk))\n",
    "    w = B[B[:,0]==d+1,1:]\n",
    "    lp += w[:,1]*np.log(b[w[:,0]])\n",
    "    nd += np.sum(w[:,1])\n",
    "perplexity = np.exp(-lp/nd)"
   ]
  }
 ],
 "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": "3.7.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}