gistfile1.txt

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "\n",
    "from math import exp, pow, sqrt\n",
    "from sklearn.metrics import roc_auc_score"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "data = pd.read_csv(\"data-logistic.csv\", header=None)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "y = data[0].values\n",
    "xs = data[[1,2]].values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "l = xs.size"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "def logistic_step(w, k, C):\n",
    "    w1, w2 = w\n",
    "    \n",
    "    sum_w1 = 0\n",
    "    sum_w2 = 0\n",
    "    for i in range(0, y.size):\n",
    "        yi = y[i]\n",
    "        xi1 = xs[i][0]\n",
    "        xi2 = xs[i][1]\n",
    "        coef = yi * ( 1 - 1 / ( 1 + exp( - yi*( w1*xi1 + w2*xi2 ))))\n",
    "        sum_w1 += xi1 * coef\n",
    "        sum_w2 += xi2 * coef\n",
    "    \n",
    "    w1_new = w1 + k*(1/l)*sum_w1 - k * C * w1\n",
    "    w2_new = w2 + k*(1/l)*sum_w2 - k * C * w2\n",
    "    \n",
    "    return (w1_new, w2_new)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "def ediff(v, w):\n",
    "    v1, v2 = v\n",
    "    w1, w2 = w\n",
    "    return sqrt(pow(v1 - w1, 2) + pow(v2 - w2, 2))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "threshold = 1e-5"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "def iterate(w0, k, C):\n",
    "    i = 0\n",
    "    w_old = w0\n",
    "    while True:\n",
    "        i += 1\n",
    "        w_new = logistic_step(w_old, k, C)\n",
    "        # print(\"i=%s, w: %s -> %s\" % (i, w_old, w_new)) \n",
    "        if ediff(w_old, w_new) < threshold:\n",
    "            break\n",
    "        w_old = w_new\n",
    "    return w_old"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "def y_calc(w, x):\n",
    "    w1, w2 = w\n",
    "    if w1 * x[0] + w2 * x[1] > 0:\n",
    "        return 1\n",
    "    else:\n",
    "        return -1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def score(w, x):\n",
    "    w1, w2 = w\n",
    "    return 1 / (1 + exp(-w1 * x[0] - w2 * x[1]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "w_reg = iterate((0,0), 0.1, 10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "y_calc_reg = np.array([ y_calc(w_reg, x) for x in xs])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "y_score_reg = np.array([ score(w_reg, x) for x in xs])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "w_no_reg = iterate((0,0), 0.1, 0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "y_calc_no_reg = np.array([ y_calc(w_no_reg, x) for x in xs])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "y_score_no_reg = np.array([ score(w_no_reg, x) for x in xs])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.93666666666666654"
      ]
     },
     "execution_count": 56,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "roc_auc_score(y, y_score_reg)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.92685714285714282"
      ]
     },
     "execution_count": 57,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "roc_auc_score(y, y_score_no_reg)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.5.1"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}