svm demo

# -*- coding: utf-8 -*-
import random
from numpy import *
def loadSet(filename):
    fr=open(filename)
    trainSet=[]
    trainLabel=[]
    for line in fr.readlines():
        currLine=line.strip().split('\t')
        trainSet.append([float(currLine[0]),float(currLine[1])])
        trainLabel.append(float(currLine[2]))
    return trainSet,trainLabel
#选择i,j。选择两个不同的alpha，SMO每次优化选择两个不同的alpha
#给定一个alpha(i),寻找另一个alpha(j)
def selectJrand(i,m):
    j=i
    while(j==i):
        j=int(random.uniform(0,m))
    return j
#调整大于H或者小于L的alph值
def clipAlpha(a,H,L):
    if a>H:
        a=H
    if a<L:
        a=L
    return a

#简化版的SMO算法
#C是松弛变量
#toler是容错率
#maxIter是最大迭代次数，外循环的最大循环数
def smoSimple(dataMatIn,classLabels,C,toler,maxIter):
    dataMatrix=mat(dataMatIn)
    labelMat=mat(classLabels).transpose()
    b=0
    m,n=shape(dataMatrix)
    alphas=mat(zeros((m,1)))
    iter=0
   
    while(iter<maxIter):
        #用于记录alp是否已经优化
        alphaPairsChanged=0
        for i in range(m):
            fxi=float(multiply(alphas,labelMat).T *\
                      ( dataMatrix *dataMatrix[i,:].T))+b
            Ei=fxi-float(labelMat[i])
           
            if((labelMat[i]*Ei<-toler)and(alphas[i]<C))or\
                ((labelMat[i]*Ei>toler)and alphas[i]>0):