def avgPerceptron(self):
### initialize variables
### initialize prediction (yHat) and actual (yStar) to (0)
### use numpy for vectors
### create a numpy array of zeroes the size of m for weight, w
#m = len(self.vocabulary)
yHat = 0
yStar = 0
count = 1
for iteration in range(self.maxIteration):
for fortune in self.featureVectorList:
### yHat and yStar are in the set {+1, -1}
yStar = sign(int(fortune[0]))
yHat = sign(self.w.dot(fortune[1]))
### if the prediction doesn't match the label,
### update the weight
if yHat != yStar:
self.w = self.w + (self.eta * yStar * np.array(fortune[1]))
self.wCache = self.wCache + (self.eta * yHat *
np.array(fortune[1]))
count += 1
numTrain = len(self.featureVectorList)
numTest = len(self.testVectorList)
trainMistakes = self.checkAccuracy(self.featureVectorList)
testMistakes = self.checkAccuracy(self.testVectorList)
self.averagedTrainingAccuracy.append(
100 - (100 * (float(trainMistakes) / numTrain)))
self.averagedTestingAccuracy.append(100 -
(100 * (float(testMistakes)) /
numTest))
'''
print 'iteration: ', iteration + 1
print 'Training mistakes: ', trainMistakes, ' Testing mistakes: ', testMistakes
print 'Training percentage: ', 100 - (100 * (float(trainMistakes)/numTrain)), '%', \
' Testing percentage: ', 100 - (100 * (float(testMistakes))/numTest), '%'
'''
'''
for vector in range(len(self.wCache)):
for element in self.wCache[vector][1]:
newWeight = float(element) / count
self.wCache[vector][1][element] = newWeight
print 'weight: ', self.wCache[vector][1]
'''
self.wCache = self.wCache / count
def checkAccuracy(self):
error = 0
for fortune in self.featureVectorList:
### yHat and yStar are in the set {+1, -1}
yStar = sign(int(fortune[0]))
yHat = sign(self.w.dot(fortune[1]))
### if the prediction doesn't match the label,
### update the weight
if yHat != yStar:
error += 1
return error
def perceptron(self):
### initialize variables
### initialize prediction (yHat) and actual (yStar) to (0)
### use numpy for vectors
### create a numpy array of zeroes the size of m for weight, w
#m = len(self.vocabulary)
yHat = 0
yStar = 0
for fortune in self.featureVectorList:
### yHat and yStar are in the set {+1, -1}
yStar = sign(int(fortune[0]))
yHat = sign(self.w.dot(fortune[1]))
### if the prediction doesn't match the label,
### update the weight
if yHat != yStar:
self.w = self.w + (self.eta * yStar * np.array(fortune[1]))
def perceptron(self):
### initialize variables
### initialize prediction (yHat) and actual (yStar) to (0)
### use numpy for vectors
### create a numpy array of zeroes the size of m for weight, w
#m = len(self.vocabulary)
for iteration in range(self.maxIteration):
mistakes = 0
yHat = 0
yStar = 0
for fortune in self.featureVectorList:
### yHat and yStar are in the set {+1, -1}
yStar = sign(int(fortune[0]))
yHat = sign(self.w.dot(fortune[1]))
### if the prediction doesn't match the label,
### update the weight
if yHat != yStar:
self.w = self.w + (self.eta * yStar * np.array(fortune[1]))
mistakes += 1
numTrain = len(self.featureVectorList)
numTest = len(self.testVectorList)
trainMistakes = self.checkAccuracy(self.featureVectorList)
testMistakes = self.checkAccuracy(self.testVectorList)
self.numberOfMistakes.append(mistakes)
self.standardTrainingAccuracy.append(
100 - (100 * (float(trainMistakes) / numTrain)))
self.standardTestingAccuracy.append(100 -
(100 * (float(testMistakes)) /
numTest))
'''
