python实现SVM【软间隔】【SMO算法】
本帖最后由 糖逗 于 2020-11-10 16:39 编辑参考书籍：《机器学习实战》
import numpy as np
from numpy import random
def loadDataSet(fileName):
dataMat = []
labelMat = []
fr = open(fileName)
for line in fr.readlines():
lineArr = line.strip().split('\t')
dataMat.append(), float(lineArr)])
labelMat.append(float(lineArr))
return dataMat, labelMat
def selectJrand(i, m):
j = i
while(j == i):
j = int(random.uniform(0, m))
return j
def clipAlpha(aj, H, L):
if aj > H:
aj = H
if L > aj:
aj = L
return aj
def smoSimple(dataMatIn, classLabels, C, toler, maxIter):
dataMatrix = np.mat(dataMatIn)
labelMat = np.mat(classLabels).transpose()
b = 0; m,n = np.shape(dataMatrix)
alphas = np.mat(np.zeros((m,1)))
iter = 0
while (iter < maxIter):
alphaPairsChanged = 0
for i in range(m):
fXi = float(np.multiply(alphas,labelMat).T*(dataMatrix*dataMatrix.T)) + b
Ei = fXi - float(labelMat)
if ((labelMat*Ei < -toler) and (alphas < C)) or ((labelMat*Ei > toler) and (alphas > 0)):
j = selectJrand(i,m)
fXj = float(np.multiply(alphas,labelMat).T*(dataMatrix*dataMatrix.T)) + b
Ej = fXj - float(labelMat)
alphaIold = alphas.copy(); alphaJold = alphas.copy();
if (labelMat != labelMat):
L = max(0, alphas - alphas)
H = min(C, C + alphas - alphas)
else:
L = max(0, alphas + alphas - C)
H = min(C, alphas + alphas)
if L==H: print("L==H"); continue
eta = 2.0 * dataMatrix*dataMatrix.T - dataMatrix*dataMatrix.T - dataMatrix*dataMatrix.T
if eta >= 0:
print("eta>=0"); continue
alphas -= labelMat*(Ei - Ej)/eta
alphas = clipAlpha(alphas,H,L)
if (abs(alphas - alphaJold) < 0.00001):
print("j not moving enough"); continue
alphas += labelMat*labelMat*(alphaJold - alphas)
b1 = b - Ei- labelMat*(alphas-alphaIold)*dataMatrix*dataMatrix.T - labelMat*(alphas-alphaJold)*dataMatrix*dataMatrix.T
b2 = b - Ej- labelMat*(alphas-alphaIold)*dataMatrix*dataMatrix.T - labelMat*(alphas-alphaJold)*dataMatrix*dataMatrix.T
if (0 < alphas) and (C > alphas):
b = b1
elif (0 < alphas) and (C > alphas):
b = b2
else:
b = (b1 + b2)/2.0
alphaPairsChanged += 1
print("iter: %d i:%d, pairs changed %d" % (iter,i,alphaPairsChanged))
if(alphaPairsChanged == 0):
iter += 1
else:
iter = 0
print("iteration number: %d" % iter)
return b,alphas
def calcWs(alphas,dataArr,classLabels):
X = np.mat(dataArr); labelMat = np.mat(classLabels).transpose()
m, n = np.shape(X)
w = np.zeros((n,1))
for i in range(m):
w += np.multiply(alphas*labelMat,X.T)
return w
dataArr, labelArr = loadDataSet(r"C:\...\testSet.txt")
b, alphas = smoSimple(dataArr, labelArr, 0.6, 0.001, 40)
ws = calcWs(alphas, dataArr, labelArr)
import seaborn as sns
import pandas as pd
import matplotlib.pyplot as plt
temp = pd.DataFrame(dataArr)
temp.columns = ["1", "2"]
temp["label"] = pd.array(labelArr)
temp["label"] = np.array(temp["label"]).astype(np.int)
xx = np.linspace(0, 10, 20)
yy = (-b - xx * ws) / ws
temp1 = pd.DataFrame()
temp1["xx"] = np.array(xx)
temp1["yy"] = np.array(yy.T)
sns.scatterplot(data = temp, x = "1", y = "2", hue = "label")
plt.plot(temp1['xx'], temp1['yy'])
本帖最后由 糖逗 于 2020-11-10 16:40 编辑
参考文献：https://www.math.pku.edu.cn/teachers/ganr/course/pr2010/Ref/platt_smoTR.pdf
参考链接：https://zhuanlan.zhihu.com/p/27662928
https://www.jianshu.com/p/eef51f939ace
页:
[1]