CS_ALearn/ailearning
1
0
Fork 0
mirror of https://github.com/apachecn/ailearning.git synced 2026-05-16 14:04:18 +08:00
Code Issues Projects Releases Wiki Activity

Merge branch 'master' of https://github.com/apachecn/MachineLearning

Browse Source
This commit is contained in:
jiangzhonglian
2017-03-27 16:44:17 +08:00
parent 60c922ea45 b33f9b0a4c
commit 79c8e672c8
1 changed files with 19 additions and 33 deletions
Download Patch File Download Diff File Expand all files Collapse all files

52
src/python/06.SVM/svmMLiA.py
View File

@@ -80,14 +80,14 @@ def smoSimple(dataMatIn, classLabels, C, toler, maxIter):
dataMatrix = mat(dataMatIn); labelMat = mat(classLabels).transpose() dataMatrix = mat(dataMatIn); labelMat = mat(classLabels).transpose()
b = 0; m,n = shape(dataMatrix) b = 0; m,n = shape(dataMatrix)
alphas = mat(zeros((m,1))) alphas = mat(zeros((m,1)))
iter = 0 iter = 0 # 没有任何alpha改变的情况下遍历数据的次数
while (iter < maxIter): while (iter < maxIter):
alphaPairsChanged = 0 alphaPairsChanged = 0 #记录alpha是否已经进行优化，每次循环时设为0，然后再对整个集合顺序遍历
for i in range(m): for i in range(m):
fXi = float(multiply(alphas,labelMat).T*(dataMatrix*dataMatrix[i,:].T)) + b fXi = float(multiply(alphas,labelMat).T*(dataMatrix*dataMatrix[i,:].T)) + b # 我们预测的类别
Ei = fXi - float(labelMat[i])#if checks if an example violates KKT conditions Ei = fXi - float(labelMat[i])#if checks if an example violates KKT conditions 误差
if ((labelMat[i]*Ei < -toler) and (alphas[i] < C)) or ((labelMat[i]*Ei > toler) and (alphas[i] > 0)): if ((labelMat[i]*Ei < -toler) and (alphas[i] < C)) or ((labelMat[i]*Ei > toler) and (alphas[i] > 0)):
j = selectJrand(i,m) j = selectJrand(i,m) # 误差很大时进行优化
fXj = float(multiply(alphas,labelMat).T*(dataMatrix*dataMatrix[j,:].T)) + b fXj = float(multiply(alphas,labelMat).T*(dataMatrix*dataMatrix[j,:].T)) + b
Ej = fXj - float(labelMat[j]) Ej = fXj - float(labelMat[j])
alphaIold = alphas[i].copy(); alphaJold = alphas[j].copy() alphaIold = alphas[i].copy(); alphaJold = alphas[j].copy()
@@ -195,18 +195,15 @@ def innerL(i, oS):
else: else:
L = max(0, oS.alphas[j] + oS.alphas[i] - oS.C) L = max(0, oS.alphas[j] + oS.alphas[i] - oS.C)
H = min(oS.C, oS.alphas[j] + oS.alphas[i]) H = min(oS.C, oS.alphas[j] + oS.alphas[i])
if L == H: print if L == H: print("L==H")
"L==H";
return 0 return 0
eta = 2.0 * oS.K[i, j] - oS.K[i, i] - oS.K[j, j] # changed for kernel eta = 2.0 * oS.K[i, j] - oS.K[i, i] - oS.K[j, j] # changed for kernel
if eta >= 0: print if eta >= 0: print("eta>=0")
"eta>=0";
return 0 return 0
oS.alphas[j] -= oS.labelMat[j] * (Ei - Ej) / eta oS.alphas[j] -= oS.labelMat[j] * (Ei - Ej) / eta
oS.alphas[j] = clipAlpha(oS.alphas[j], H, L) oS.alphas[j] = clipAlpha(oS.alphas[j], H, L)
updateEk(oS, j) # added this for the Ecache updateEk(oS, j) # added this for the Ecache
if (abs(oS.alphas[j] - alphaJold) < 0.00001): print if (abs(oS.alphas[j] - alphaJold) < 0.00001): print("j not moving enough")
"j not moving enough";
return 0 return 0
oS.alphas[i] += oS.labelMat[j] * oS.labelMat[i] * (alphaJold - oS.alphas[j]) # update i by the same amount as j oS.alphas[i] += oS.labelMat[j] * oS.labelMat[i] * (alphaJold - oS.alphas[j]) # update i by the same amount as j
updateEk(oS, i) # added this for the Ecache #the update is in the oppostie direction updateEk(oS, i) # added this for the Ecache #the update is in the oppostie direction
@@ -235,22 +232,19 @@ def smoP(dataMatIn, classLabels, C, toler, maxIter, kTup=('lin', 0)): # full Pl
if entireSet: # go over all if entireSet: # go over all
for i in range(oS.m): for i in range(oS.m):
alphaPairsChanged += innerL(i, oS) alphaPairsChanged += innerL(i, oS)
print print("fullSet, iter: %d i:%d, pairs changed %d" % (iter, i, alphaPairsChanged))
"fullSet, iter: %d i:%d, pairs changed %d" % (iter, i, alphaPairsChanged)
iter += 1 iter += 1
else: # go over non-bound (railed) alphas else: # go over non-bound (railed) alphas
nonBoundIs = nonzero((oS.alphas.A > 0) * (oS.alphas.A < C))[0] nonBoundIs = nonzero((oS.alphas.A > 0) * (oS.alphas.A < C))[0]
for i in nonBoundIs: for i in nonBoundIs:
alphaPairsChanged += innerL(i, oS) alphaPairsChanged += innerL(i, oS)
print print("non-bound, iter: %d i:%d, pairs changed %d" % (iter, i, alphaPairsChanged))
"non-bound, iter: %d i:%d, pairs changed %d" % (iter, i, alphaPairsChanged)
iter += 1 iter += 1
if entireSet: if entireSet:
entireSet = False # toggle entire set loop entireSet = False # toggle entire set loop
elif (alphaPairsChanged == 0): elif (alphaPairsChanged == 0):
entireSet = True entireSet = True
print print("iteration number: %d" % iter)
"iteration number: %d" % iter
return oS.b, oS.alphas return oS.b, oS.alphas
@@ -272,16 +266,14 @@ def testRbf(k1=1.3):
svInd = nonzero(alphas.A > 0)[0] svInd = nonzero(alphas.A > 0)[0]
sVs = datMat[svInd] # get matrix of only support vectors sVs = datMat[svInd] # get matrix of only support vectors
labelSV = labelMat[svInd]; labelSV = labelMat[svInd];
print print("there are %d Support Vectors" % shape(sVs)[0])
"there are %d Support Vectors" % shape(sVs)[0]
m, n = shape(datMat) m, n = shape(datMat)
errorCount = 0 errorCount = 0
for i in range(m): for i in range(m):
kernelEval = kernelTrans(sVs, datMat[i, :], ('rbf', k1)) kernelEval = kernelTrans(sVs, datMat[i, :], ('rbf', k1))
predict = kernelEval.T * multiply(labelSV, alphas[svInd]) + b predict = kernelEval.T * multiply(labelSV, alphas[svInd]) + b
if sign(predict) != sign(labelArr[i]): errorCount += 1 if sign(predict) != sign(labelArr[i]): errorCount += 1
print print("the training error rate is: %f" % (float(errorCount) / m))
"the training error rate is: %f" % (float(errorCount) / m)
dataArr, labelArr = loadDataSet('testSetRBF2.txt') dataArr, labelArr = loadDataSet('testSetRBF2.txt')
errorCount = 0 errorCount = 0
datMat = mat(dataArr); datMat = mat(dataArr);
@@ -291,8 +283,7 @@ def testRbf(k1=1.3):
kernelEval = kernelTrans(sVs, datMat[i, :], ('rbf', k1)) kernelEval = kernelTrans(sVs, datMat[i, :], ('rbf', k1))
predict = kernelEval.T * multiply(labelSV, alphas[svInd]) + b predict = kernelEval.T * multiply(labelSV, alphas[svInd]) + b
if sign(predict) != sign(labelArr[i]): errorCount += 1 if sign(predict) != sign(labelArr[i]): errorCount += 1
print print("the test error rate is: %f" % (float(errorCount) / m))
"the test error rate is: %f" % (float(errorCount) / m)
def img2vector(filename): def img2vector(filename):
@@ -338,8 +329,7 @@ def testDigits(kTup=('rbf', 10)):
kernelEval = kernelTrans(sVs, datMat[i, :], kTup) kernelEval = kernelTrans(sVs, datMat[i, :], kTup)
predict = kernelEval.T * multiply(labelSV, alphas[svInd]) + b predict = kernelEval.T * multiply(labelSV, alphas[svInd]) + b
if sign(predict) != sign(labelArr[i]): errorCount += 1 if sign(predict) != sign(labelArr[i]): errorCount += 1
print print("the training error rate is: %f" % (float(errorCount) / m))
"the training error rate is: %f" % (float(errorCount) / m)
dataArr, labelArr = loadImages('testDigits') dataArr, labelArr = loadImages('testDigits')
errorCount = 0 errorCount = 0
datMat = mat(dataArr); datMat = mat(dataArr);
@@ -349,8 +339,7 @@ def testDigits(kTup=('rbf', 10)):
kernelEval = kernelTrans(sVs, datMat[i, :], kTup) kernelEval = kernelTrans(sVs, datMat[i, :], kTup)
predict = kernelEval.T * multiply(labelSV, alphas[svInd]) + b predict = kernelEval.T * multiply(labelSV, alphas[svInd]) + b
if sign(predict) != sign(labelArr[i]): errorCount += 1 if sign(predict) != sign(labelArr[i]): errorCount += 1
print print("the test error rate is: %f" % (float(errorCount) / m))
"the test error rate is: %f" % (float(errorCount) / m)
'''#######******************************** '''#######********************************
@@ -416,18 +405,15 @@ def innerLK(i, oS):
else: else:
L = max(0, oS.alphas[j] + oS.alphas[i] - oS.C) L = max(0, oS.alphas[j] + oS.alphas[i] - oS.C)
H = min(oS.C, oS.alphas[j] + oS.alphas[i]) H = min(oS.C, oS.alphas[j] + oS.alphas[i])
if L == H: print if L == H: print("L==H")
"L==H";
return 0 return 0
eta = 2.0 * oS.X[i, :] * oS.X[j, :].T - oS.X[i, :] * oS.X[i, :].T - oS.X[j, :] * oS.X[j, :].T eta = 2.0 * oS.X[i, :] * oS.X[j, :].T - oS.X[i, :] * oS.X[i, :].T - oS.X[j, :] * oS.X[j, :].T
if eta >= 0: print if eta >= 0: print("eta>=0")
"eta>=0";
return 0 return 0
oS.alphas[j] -= oS.labelMat[j] * (Ei - Ej) / eta oS.alphas[j] -= oS.labelMat[j] * (Ei - Ej) / eta
oS.alphas[j] = clipAlpha(oS.alphas[j], H, L) oS.alphas[j] = clipAlpha(oS.alphas[j], H, L)
updateEk(oS, j) # added this for the Ecache updateEk(oS, j) # added this for the Ecache
if (abs(oS.alphas[j] - alphaJold) < 0.00001): print if (abs(oS.alphas[j] - alphaJold) < 0.00001): print("j not moving enough")
"j not moving enough";
return 0 return 0
oS.alphas[i] += oS.labelMat[j] * oS.labelMat[i] * (alphaJold - oS.alphas[j]) # update i by the same amount as j oS.alphas[i] += oS.labelMat[j] * oS.labelMat[i] * (alphaJold - oS.alphas[j]) # update i by the same amount as j
updateEk(oS, i) # added this for the Ecache #the update is in the oppostie direction updateEk(oS, i) # added this for the Ecache #the update is in the oppostie direction