label propagation.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# 数据读入和整理\n",
    "iris = pd.read_csv('iris.csv')\n",
    "X = iris.ix[:,:4].values\n",
    "mapdict ={k:v for v, k in enumerate(iris.Species.unique())}\n",
    "y = iris.Species.map(mapdict).values\n",
    "n  = len(y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# 切分数据\n",
    "train_index = np.random.choice(n, 0.6*n, replace = False)\n",
    "test_index = np.setdiff1d(np.arange(n),train_index)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# 计算权重矩阵\n",
    "sigma = X.var(axis = 0)\n",
    "weights = np.zeros((n,n))\n",
    "\n",
    "def weight_func(ind1,ind2,X=X,sigma=sigma):\n",
    "    return np.exp(-np.sum((X[ind1]-X[ind2])**2/sigma))\n",
    "\n",
    "for i in range(n):\n",
    "    for j in range(n):\n",
    "        weights[i,j] = weight_func(i,j)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "##  标准化为转移矩阵\n",
    "t = weights/weights.sum(axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# y转换形式\n",
    "y_m = np.zeros((n, len(np.unique(y))))\n",
    "for i in range(n):\n",
    "    y_m[i,y[i]] = 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "## unlabel初始化, label记住\n",
    "y_m[test_index] = np.random.random(y_m[test_index].shape)\n",
    "clamp = y_m[train_index]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "## 迭代计算\n",
    "iter_n = 50\n",
    "for _ in range(iter_n):\n",
    "    y_m = t.dot(y_m)  # LP\n",
    "    y_m = (y_m.T/y_m.sum(axis=1)).T # normalize\n",
    "    y_m[train_index] = clamp  # clamp"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.93333333333333335"
      ]
     },
     "execution_count": 38,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#  预测准确率\n",
    "predict = y_m[test_index].argmax(axis=1)\n",
    "np.sum(y[test_index] == predict)/float(len(predict))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "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": "2.7.11"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}