smo

还在更新中


import numpy as np
import random

# build auxiliary function

class optstruct:
    def __init__(self, datax, datay, C, toler):
        self.x = datax
        self.label = datay
        self.c = C
        self.tol = toler
        self.m = datax.shape[0]  # how many samples are there
        self.alphas = np.mat(np.zeros((self.m, 1)))  # initialize alphas
        self.b = 0  # initialize b
        self.ercache = np.mat(np.zeros((self.m, 2)))

def calcek(os,k):
    fxk=float(np.multiply(os.alphas,os.label).T*(os.x*os.x[k,:].T))+os.b # fx=wx+b,w=alphasi*yi*xi sum from n
    Ek=fxk-float(os.label[k])  # Ek=yi-labeli
    return Ek



def pickJ(i,os,Ei):
    maxk=-1; maxdeltaE=0;Ej=0
    os.ercache[i]=[1,Ei]
    valerchache=np.nonzero(os.ercache[:,0].A)[0] # work for .A matrix convert array  # index value
    if (len(valerchache))>1: # if having value , do it
        for k in valerchache: # traversal index
            if k==i:
                continue
            Ek=calcek(os,k)
            deltas=abs(Ei-Ek)
            if deltas>maxdeltaE:  # heuristic algorithm
                maxk=k
                maxdeltaE=deltas
                Ej=Ek
        return maxk,Ej
    else:
        j=i
        while(j==i):
            j=int(random.uniform(0,os.m))
        Ej=calcek(os,j)
        return j,Ej
def updateEk(os,k):
    Ek=calcek(os,k)
    os.ercache[k]=[1,Ek]