Kernel Meam Matching.ipynb

{
 "metadata": {
  "name": "",
  "signature": "sha256:a354edc51096828fb2fb15fa7e38e3c614657c3d00306805c0ed85a28b343add"
 },
 "nbformat": 3,
 "nbformat_minor": 0,
 "worksheets": [
  {
   "cells": [
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "import os\n",
      "import sys\n",
      "import numpy as np \n",
      "import math\n",
      "import matplotlib.pyplot as plt\n",
      "from cvxopt import matrix, solvers\n",
      "%matplotlib inline"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 1
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "# an implementation of Kernel Mean Matchin\n",
      "# referenres:\n",
      "#  1. Gretton, Arthur, et al. \"Covariate shift by kernel mean matching.\" Dataset shift in machine learning 3.4 (2009): 5.\n",
      "#  2. Huang, Jiayuan, et al. \"Correcting sample selection bias by unlabeled data.\" Advances in neural information processing systems. 2006.\n",
      "def kernel_mean_matching(X, Z, kern='lin', B=1.0, eps=None):\n",
      "    nx = X.shape[0]\n",
      "    nz = Z.shape[0]\n",
      "    if eps == None:\n",
      "        eps = B/math.sqrt(nz)\n",
      "    if kern == 'lin':\n",
      "        K = np.dot(Z, Z.T)\n",
      "        kappa = np.sum(np.dot(Z, X.T)*float(nz)/float(nx),axis=1)\n",
      "    elif kern == 'rbf':\n",
      "        K = compute_rbf(Z,Z)\n",
      "        kappa = np.sum(compute_rbf(Z,X),axis=1)*float(nz)/float(nx)\n",
      "    else:\n",
      "        raise ValueError('unknown kernel')\n",
      "        \n",
      "    K = matrix(K)\n",
      "    kappa = matrix(kappa)\n",
      "    G = matrix(np.r_[np.ones((1,nz)), -np.ones((1,nz)), np.eye(nz), -np.eye(nz)])\n",
      "    h = matrix(np.r_[nz*(1+eps), nz*(eps-1), B*np.ones((nz,)), np.zeros((nz,))])\n",
      "    \n",
      "    sol = solvers.qp(K, -kappa, G, h)\n",
      "    coef = np.array(sol['x'])\n",
      "    return coef\n",
      "\n",
      "def compute_rbf(X, Z, sigma=1.0):\n",
      "    K = np.zeros((X.shape[0], Z.shape[0]), dtype=float)\n",
      "    for i, vx in enumerate(X):\n",
      "        K[i,:] = np.exp(-np.sum((vx-Z)**2, axis=1)/(2.0*sigma))\n",
      "    return K"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 196
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "x = 11*np.random.random(200)- 6.0\n",
      "y = x**2 + 10*np.random.random(200) - 5\n",
      "Z = np.c_[x, y]\n",
      "\n",
      "x = 2*np.random.random(10) - 6.0\n",
      "y = x**2 + 10*np.random.random(10) - 5\n",
      "X = np.c_[x, y]\n",
      "\n",
      "coef = kernel_mean_matching(X, Z, kern='rbf', B=10)\n",
      "\n",
      "plt.close()\n",
      "plt.figure()\n",
      "plt.scatter(Z[:,0], Z[:,1], color='black', marker='x')\n",
      "plt.scatter(X[:,0], X[:,1], color='red')\n",
      "plt.scatter(Z[:,0], Z[:,1], color='green', s=coef*10, alpha=0.5)\n",
      "\n",
      "np.sum(coef > 1e-2)"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "     pcost       dcost       gap    pres   dres\n",
        " 0: -2.3208e+03 -2.7720e+05  3e+05  6e-02  3e-14\n",
        " 1: -1.4630e+03 -1.0744e+05  1e+05  2e-02  9e-15\n",
        " 2: -1.5771e+03 -3.0576e+04  3e+04  4e-03  3e-15\n",
        " 3: -1.5352e+03 -2.8400e+04  3e+04  2e-03  1e-15\n",
        " 4: -1.8948e+03 -1.9183e+04  2e+04  1e-03  8e-16\n",
        " 5: -2.1878e+03 -9.5321e+03  7e+03  3e-04  1e-15"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        " 6: -2.2752e+03 -5.4322e+03  3e+03  2e-05  1e-15"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        " 7: -2.5074e+03 -3.1354e+03  6e+02  6e-16  3e-16\n",
        " 8: -2.6198e+03 -2.7294e+03  1e+02  2e-16  2e-16\n",
        " 9: -2.6468e+03 -2.6620e+03  2e+01  7e-16  2e-16"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "10: -2.6510e+03 -2.6536e+03  3e+00  2e-16  2e-16"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "11: -2.6518e+03 -2.6521e+03  3e-01  1e-15  1e-16"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "12: -2.6519e+03 -2.6520e+03  4e-02  3e-16  2e-16"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "13: -2.6520e+03 -2.6520e+03  5e-03  5e-16  1e-16"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "14: -2.6520e+03 -2.6520e+03  7e-04  3e-16  2e-16"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "Optimal solution found.\n"
       ]
      },
      {
       "metadata": {},
       "output_type": "pyout",
       "prompt_number": 213,
       "text": [
        "24"
       ]
      },
      {
       "metadata": {},
       "output_type": "display_data",
       "png": "iVBORw0KGgoAAAANSUhEUgAAAXYAAAD9CAYAAACoXlzKAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nGUzYGQwGo5XBhJ3BYDBaGf8PZKwdS0PfKB4AAAAASUVORK5CYII=\n",
       "text": [
        "<matplotlib.figure.Figure at 0xcd1f090>"
       ]
      }
     ],
     "prompt_number": 213
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "coef"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "metadata": {},
       "output_type": "pyout",
       "prompt_number": 192,
       "text": [
        "array([[  1.81320645e-03],\n",
        "       [  1.82585202e-03],\n",
        "       [  1.17411692e-06],\n",
        "       [  2.11298174e-03],\n",
        "       [  2.39296443e-03],\n",
        "       [  1.45552789e-06],\n",
        "       [  1.20208348e-05],\n",
        "       [  1.78416405e-03],\n",
        "       [  4.56741278e+00],\n",
        "       [  2.65791856e-03],\n",
        "       [  1.28088497e-05],\n",
        "       [  1.29012304e-06],\n",
        "       [  1.95755860e-03],\n",
        "       [  2.03076280e-03],\n",
        "       [  2.06754603e-06],\n",
        "       [  1.77122399e-03],\n",
        "       [  1.71002555e-03],\n",
        "       [  9.06115192e-05],\n",
        "       [  6.73450267e-06],\n",
        "       [  1.76576042e-03],\n",
        "       [  8.44093393e+00],\n",
        "       [  6.09962758e-07],\n",
        "       [  2.09933807e-06],\n",
        "       [  1.80911719e-03],\n",
        "       [  4.76409659e-06],\n",
        "       [  3.55591415e-06],\n",
        "       [  3.43323852e+00],\n",
        "       [  3.55640762e-04],\n",
        "       [  2.32053571e-03],\n",
        "       [  1.52323916e-06],\n",
        "       [  3.36143300e-04],\n",
        "       [  1.86499023e-03],\n",
        "       [  1.00836003e-06],\n",
        "       [  6.53381636e+00],\n",
        "       [  8.68127518e-06],\n",
        "       [  8.43009148e-07],\n",
        "       [  3.21119836e-03],\n",
        "       [  1.81540613e-06],\n",
        "       [  2.35209364e-03],\n",
        "       [  9.99999964e+00],\n",
        "       [  1.11606747e-04],\n",
        "       [  1.90419817e-03],\n",
        "       [  2.01197574e-03],\n",
        "       [  8.03171129e+00],\n",
        "       [  3.27615611e+00],\n",
        "       [  2.13659841e-03],\n",
        "       [  5.63681595e-07],\n",
        "       [  4.79178418e-06],\n",
        "       [  1.09034531e-06],\n",
        "       [  2.96114714e-03],\n",
        "       [  4.23827000e+00],\n",
        "       [  2.69138526e-04],\n",
        "       [  2.07914289e-06],\n",
        "       [  2.14332343e-06],\n",
        "       [  1.14072451e-06],\n",
        "       [  6.64976392e+00],\n",
        "       [  2.68480456e-03],\n",
        "       [  7.11368940e-07],\n",
        "       [  7.66218122e-07],\n",
        "       [  2.88662603e-03],\n",
        "       [  7.51321782e-07],\n",
        "       [  1.82036001e-03],\n",
        "       [  6.04652574e-04],\n",
        "       [  9.99999920e+00],\n",
        "       [  9.99999983e+00],\n",
        "       [  2.24528564e-06],\n",
        "       [  9.89562506e-07],\n",
        "       [  2.56519530e-03],\n",
        "       [  1.45878922e-06],\n",
        "       [  1.81560855e-03],\n",
        "       [  7.67120870e-07],\n",
        "       [  4.48252320e-03],\n",
        "       [  6.50066036e-03],\n",
        "       [  5.89658680e-07],\n",
        "       [  5.48930840e-07],\n",
        "       [  4.63988938e-03],\n",
        "       [  1.01203863e-06],\n",
        "       [  7.87098260e-07],\n",
        "       [  1.91750640e-03],\n",
        "       [  3.58354678e-05],\n",
        "       [  5.28767165e-06],\n",
        "       [  2.37922297e-03],\n",
        "       [  2.32794791e-03],\n",
        "       [  1.88537264e-03],\n",
        "       [  5.70039147e+00],\n",
        "       [  6.60781282e+00],\n",
        "       [  2.21076778e-03],\n",
        "       [  5.68543192e-07],\n",
        "       [  2.39357468e-03],\n",
        "       [  1.36923384e-06],\n",
        "       [  1.78794735e-03],\n",
        "       [  2.57954230e-06],\n",
        "       [  5.01060752e+00],\n",
        "       [  3.59648689e-03],\n",
        "       [  1.21130521e-04],\n",
        "       [  2.22470710e-06],\n",
        "       [  4.14758353e-06],\n",
        "       [  3.31151252e-06],\n",
        "       [  7.66552104e-05],\n",
        "       [  1.93407737e-04],\n",
        "       [  7.41072609e-04],\n",
        "       [  2.41951255e-03],\n",
        "       [  3.68446812e-06],\n",
        "       [  9.99999929e+00],\n",
        "       [  2.41452763e-06],\n",
        "       [  2.10319095e-03],\n",
        "       [  2.15606603e-03],\n",
        "       [  2.35586972e-06],\n",
        "       [  9.99999647e+00],\n",
        "       [  2.13395461e-03],\n",
        "       [  5.06780011e-01],\n",
        "       [  2.27908180e-03],\n",
        "       [  6.81091004e-06],\n",
        "       [  2.87789032e-06],\n",
        "       [  2.38696938e-03],\n",
        "       [  2.06656483e-03],\n",
        "       [  9.99999993e+00],\n",
        "       [  5.24945447e-07],\n",
        "       [  4.15070160e-01],\n",
        "       [  1.71794966e-03],\n",
        "       [  4.71236660e-03],\n",
        "       [  1.97851477e-03],\n",
        "       [  2.06594702e-03],\n",
        "       [  2.30664235e-03],\n",
        "       [  9.31918993e-06],\n",
        "       [  2.76979343e-03],\n",
        "       [  1.36681937e-07],\n",
        "       [  1.83197084e-03],\n",
        "       [  1.87227032e-03],\n",
        "       [  9.99999956e+00],\n",
        "       [  2.37179432e-03],\n",
        "       [  1.81673589e-03],\n",
        "       [  1.80843719e-06],\n",
        "       [  2.74720790e-03],\n",
        "       [  1.93776096e-03],\n",
        "       [  8.71965314e+00],\n",
        "       [  9.77298862e-01],\n",
        "       [  2.21911853e-03],\n",
        "       [  8.26068814e+00],\n",
        "       [  1.85128069e-03],\n",
        "       [  1.38317166e-01],\n",
        "       [  1.79508297e-03],\n",
        "       [  2.13176645e-04],\n",
        "       [  2.52548876e-06],\n",
        "       [  6.34089527e+00],\n",
        "       [  9.51454306e-07],\n",
        "       [  2.41416338e-03],\n",
        "       [  1.83191536e-03],\n",
        "       [  8.52152719e-07],\n",
        "       [  8.40545960e-04],\n",
        "       [  1.76334630e-03],\n",
        "       [  8.85226154e-06],\n",
        "       [  4.36567894e-04],\n",
        "       [  2.48194926e-03],\n",
        "       [  1.96093722e-03],\n",
        "       [  9.16255370e-07],\n",
        "       [  2.92916427e-03],\n",
        "       [  3.55383983e-06],\n",
        "       [  3.19048140e-06],\n",
        "       [  5.48081884e-05],\n",
        "       [  4.12492437e-05],\n",
        "       [  1.83636820e-03],\n",
        "       [  8.06248133e-06],\n",
        "       [  1.68510381e-03],\n",
        "       [  1.35628300e-06],\n",
        "       [  3.31178584e+00],\n",
        "       [  1.71688097e-03],\n",
        "       [  1.87677364e-03],\n",
        "       [  1.85357164e-03],\n",
        "       [  1.36547209e-01],\n",
        "       [  1.76242258e-03],\n",
        "       [  1.00447944e-05],\n",
        "       [  9.42665928e-07],\n",
        "       [  9.99999992e+00],\n",
        "       [  4.00093914e-05],\n",
        "       [  9.99999985e+00],\n",
        "       [  5.18558892e-07],\n",
        "       [  1.83806773e-03],\n",
        "       [  2.72956021e-03],\n",
        "       [  1.78821994e-03],\n",
        "       [  1.42839714e-06],\n",
        "       [  1.83643458e-03],\n",
        "       [  8.74678528e-07],\n",
        "       [  4.34950733e-04],\n",
        "       [  5.16073034e-05],\n",
        "       [  1.86010060e-03],\n",
        "       [  2.22547113e-03],\n",
        "       [  5.72539201e-06],\n",
        "       [  1.74729643e-03],\n",
        "       [  8.27271466e-07],\n",
        "       [  1.76473771e-03],\n",
        "       [  1.71349282e-03],\n",
        "       [  4.53368261e-06],\n",
        "       [  9.99999955e+00],\n",
        "       [  2.60591650e-05],\n",
        "       [  1.83064920e-03],\n",
        "       [  1.83194766e-03],\n",
        "       [  1.85693846e-03],\n",
        "       [  5.30368387e-05],\n",
        "       [  8.48659418e-06]])"
       ]
      }
     ],
     "prompt_number": 192
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [],
     "language": "python",
     "metadata": {},
     "outputs": []
    }
   ],
   "metadata": {}
  }
 ]
}