def runNaiveBayesMultiTest(datasize=100):
trainsampleinput = samples.readInput('digitdata/trainingimages',
DIGIT_DATUM_WIDTH, DIGIT_DATUM_HEIGHT)
trainlabels = samples.readLabels('digitdata/traininglabels', 'multi')
datasize = (datasize * len(trainsampleinput)) / 100
labelpos = getLabelPosition(trainlabels[:datasize], 'multi')
trainfeatures = samples.featureExtraction(trainsampleinput[:datasize], 4,
4)
# getLabelDetails(trainlabels, 'multi', datasize)
trainprobs = np.loadtxt("naiveprobmulti" + str(datasize) + ".txt")
trainprobstrans = np.transpose(trainprobs)
testsampleinput = samples.readInput('digitdata/testimages',
DIGIT_DATUM_WIDTH, DIGIT_DATUM_HEIGHT)
testlabels = samples.readLabels('digitdata/testlabels', 'multi')
testfeatures = samples.featureExtraction(testsampleinput, 4, 4)
count = 0
errorcount = 0
for sample in testfeatures:
# profile.runctx('print(naiveBayesTrain(labelpos, trainfeatures, sample, trainprobstrans)); print()',globals(),{'labelpos': labelpos, 'trainfeatures': trainfeatures, 'sample': sample, 'trainprobstrans':trainprobstrans[2] },)
prediction = naiveBayesTrain(labelpos, trainfeatures, sample,
trainprobstrans[2])
# print "predicted %s actual %s" % (prediction, testlabels[count])
if prediction != testlabels[count]:
errorcount += 1
count += 1
prederror = float(errorcount) / len(testlabels)
print "Prediction accuracy for Naive Bayes MultiClass %s." % (
(1 - prederror) * 100)
def runNaiveBayesBinaryTest(datasize):
trainsampleinput = samples.readInput('facedata/facedatatrain',
FACE_DATUM_WIDTH, FACE_DATUM_HEIGHT)
datasize = (datasize * len(trainsampleinput)) / 100
trainlabels = samples.readLabels('facedata/facedatatrainlabels', 'binary')
labelpos = getLabelPosition(trainlabels[:datasize], 'binary')
trainfeatures = samples.featureExtraction(trainsampleinput[:datasize], 6,
7)
getLabelDetails(trainlabels, 'binary', datasize)
trainprobs = np.loadtxt("naiveprobbinary" + str(datasize) + ".txt")
trainprobstrans = np.transpose(trainprobs)
testsampleinput = samples.readInput('facedata/facedatatest',
FACE_DATUM_WIDTH, FACE_DATUM_HEIGHT)
testlabels = samples.readLabels('facedata/facedatatestlabels', 'binary')
testfeatures = samples.featureExtraction(testsampleinput, 6, 7)
count = 0
errorcount = 0
for sample in testfeatures:
# profile.runctx('print(naiveBayesTrain(labelpos, trainfeatures, sample, trainprobstrans)); print()',globals(),{'labelpos': labelpos, 'trainfeatures': trainfeatures, 'sample': sample, 'trainprobstrans':trainprobstrans[2] },)
prediction = naiveBayesTrain(labelpos, trainfeatures, sample,
trainprobstrans[2])
# print "predicted %s actual %s" % (prediction, testlabels[count])
if prediction != testlabels[count]:
errorcount += 1
count += 1
prederror = float(errorcount) / len(testlabels)
print "Prediction accuracy for Naive Bayes Binary Class data %s." % (
(1 - prederror) * 100)
def runKnnTesterMulti(datasize=100):
print 'Running KNN MultiClass for datasize %s' % (datasize)
testsampleinput = samples.readInput('digitdata/testimages', DIGIT_DATUM_WIDTH, DIGIT_DATUM_HEIGHT)
testactuallabels = samples.readLabels('digitdata/testlabels', 'multi')
testfeatures = knnfeatureExtract(testsampleinput[:datasize])
#testfeatures = samples.featureExtraction(testsampleinput, 4, 4)
trainsampleinput = samples.readInput('digitdata/trainingimages', DIGIT_DATUM_WIDTH, DIGIT_DATUM_HEIGHT)
datasize = (datasize*len(trainsampleinput))/100
trainactuallabels = samples.readLabels('digitdata/traininglabels', 'multi')
#trainfeatures = samples.featureExtraction(trainsampleinput, 4, 4)
trainfeatures = knnfeatureExtract(trainsampleinput[:datasize])
testknn(testfeatures, testactuallabels, trainfeatures, trainactuallabels[:datasize], 5)
def runKnnTesterBinary(datasize=100):
print 'Running KNN BinaryClass for datasize %s' % (datasize)
testsampleinput = samples.readInput('facedata/facedatatest', FACE_DATUM_WIDTH, FACE_DATUM_HEIGHT)
testactuallabels = samples.readLabels('facedata/facedatatestlabels', 'binary','knn')
testfeatures = knnfeatureExtract(testsampleinput)
#testfeatures = samples.featureExtraction(testsampleinput, 6, 7)
trainsampleinput = samples.readInput('facedata/facedatatrain', FACE_DATUM_WIDTH, FACE_DATUM_HEIGHT)
datasize = (datasize*len(trainsampleinput))/100
trainactuallabels = samples.readLabels('facedata/facedatatrainlabels', 'binary', 'knn')
trainfeatures = knnfeatureExtract(trainsampleinput[:datasize])
#trainfeatures = samples.featureExtraction(trainsampleinput, 6, 7)
print('Running KNN binaryclass \n')
testknn(testfeatures, testactuallabels, trainfeatures, trainactuallabels[:datasize], 5)
def runPercepMultiValidator(datasize=100):
sampleinput = samples.readInput('digitdata/testimages', DIGIT_DATUM_WIDTH,
DIGIT_DATUM_HEIGHT)
actuallabels = samples.readLabels('digitdata/testlabels', 'multi')
features = samples.featureExtraction(sampleinput, 4, 4)
sampleinputtain = samples.readInput('digitdata/trainingimages',
DIGIT_DATUM_WIDTH, DIGIT_DATUM_HEIGHT)
datasize = (datasize * len(sampleinputtain)) / 100
weights = np.loadtxt('percepmultiweights' + str(datasize) + '.txt')
count = 0
errorcount = 0
for label in features:
guess = guessPercep(label, weights)
#print "predicted %s actual %s" % (guess, actuallabels[count])
if guess != int(actuallabels[count]):
errorcount += 1
count += 1
prederror = float(errorcount) / len(actuallabels)
#prederror = errorcount/len(actuallabels)
print "Prediction accuracy for perceptron multi class %s." % (
(1 - prederror) * 100)
def runPercepBinaryTrain(maxiter, datasize):
sampleinput = samples.readInput('facedata/facedatatrain', FACE_DATUM_WIDTH,
FACE_DATUM_HEIGHT)
datasize = (datasize * len(sampleinput)) / 100
actuallabels = samples.readLabels('facedata/facedatatrainlabels', 'binary')
features = samples.featureExtraction(sampleinput[:datasize], 6, 7)
featlength = 0
for label in features[:datasize]:
featlength = len(label)
break
weights = getInitialWeights('binary', featlength)
trainPerc(weights, features, actuallabels[:datasize], maxiter, 'binary',
datasize)
def runPercepMultiClassTrain(maxiter, datasize):
sampleinput = samples.readInput('digitdata/trainingimages',
DIGIT_DATUM_WIDTH, DIGIT_DATUM_HEIGHT)
datasize = (datasize * len(sampleinput)) / 100
actuallabels = samples.readLabels('digitdata/traininglabels', 'multi')
features = samples.featureExtraction(sampleinput[:datasize], 4, 4)
featlength = 0
for label in features[:datasize]:
featlength = len(label)
break
weights = getInitialWeights('multi', featlength)
trainPerc(weights, features, actuallabels[:datasize], maxiter, 'multi',
datasize)
def runPercepBinaryValidator(datasize=100):
sampleinput = samples.readInput('facedata/facedatatest', FACE_DATUM_WIDTH,
FACE_DATUM_HEIGHT)
actuallabels = samples.readLabels('facedata/facedatatestlabels', 'binary')
features = samples.featureExtraction(sampleinput, 6, 7)
sampleinputtrain = samples.readInput('facedata/facedatatrain',
FACE_DATUM_WIDTH, FACE_DATUM_HEIGHT)
datasize = (datasize * len(sampleinputtrain)) / 100
weights = np.loadtxt('percepbinaryweights' + str(datasize) + '.txt')
count = 0
errorcount = 0
for label in features:
guess = guessPercep(label, weights)
#print "predicted %s actual %s" % (guess, actuallabels[count])
if guess != int(actuallabels[count]):
errorcount += 1
count += 1
prederror = float(errorcount) / len(actuallabels)
#prederror = errorcount/len(actuallabels)
print "Prediction accuracy for perceptron Binry Class %s." % (
(1 - prederror) * 100)