def Run_Q3_3_2_sp():
print '*********************************************************************************'
print "//Question-3.3.2 (Simple Perceptron-single pass) Report on Adult Data -----------------------------"
print '*********************************************************************************'
[XData, YData, FSize, SeenWt] = func.parseInfo(trainFileHandle)
maxVecLen = max(FSize)
margin = 0
for i in range(1, 2):
for j in LRateList:
[Bias, Wvec,
LearningMistake] = func.Perceptron(j, 1, XData, YData, maxVecLen,
i, margin, 0, SeenWt)
mistakeCount = func.TestPerceptron(XData, YData, Wvec, Bias,
maxVecLen, margin)
TrainAccuracy = (
(float(len(XData) - mistakeCount)) / len(XData)) * 100
print 'Using Learning Rate = ', j, ' number of epochs = Sinlge Pass, Initialization = random with mean 0 and sd 0.01'
# print '\t Learned Weight Vector = ', Wvec
print '\t Learned Bias = ', Bias
print '\t Mistakes during Learning = ', LearningMistake
print '\t Learning Accuracy = ', TrainAccuracy, '%\n'
if (test_index != -1):
[testXData,
testYData] = func.parseInfoTest(testFilehandle, maxVecLen)
testMistake = func.TestPerceptron(testXData, testYData, Wvec,
Bias, maxVecLen, margin)
TestAccuracy = ((float(len(testXData) - testMistake)) /
len(testXData)) * 100
print '\t Test Accuracy = ', TestAccuracy, '%\n'
print "//End of Question-3.3.2(simple Perceptron-single pass) Report ------------------------------------\n\n\n"
def Run_Q3_grad_with_Shuffle():
print '*********************************************************************************'
print "//Question-3: For Grads (Aggressive Perceptron-batch 3,4,5) With Shuffle Report on Adult Data Set -------------"
print '*********************************************************************************'
[XData, YData, FSize, SeenWt] = func.parseInfo(trainFileHandle)
maxVecLen = max(FSize)
for i in range(3, 6):
for k in MarginList:
[Bias, Wvec, LearningMistake
] = func.AggressivePerceptron(1, XData, YData, maxVecLen, i, k, 1,
SeenWt)
mistakeCount = func.TestMarginPerceptron(XData, YData, Wvec, Bias,
maxVecLen, k)
TrainAccuracy = (
(float(len(XData) - mistakeCount)) / len(XData)) * 100
print 'Using Dynamic Learning rate, number of epochs = ', i, ', Margin = ', k, ', Initialization = random with mean 0 and sd 0.01, with Shuffle'
#print '\t Learned Weight Vector = ', Wvec
print '\t Learned Bias = ', Bias
print '\t Mistakes during Learning = ', LearningMistake
print '\t Learning Accuracy = ', TrainAccuracy, '%\n'
if (test_index != -1):
[testXData,
testYData] = func.parseInfoTest(testFilehandle, maxVecLen)
testMistake = func.TestMarginPerceptron(
testXData, testYData, Wvec, Bias, maxVecLen, k)
TestAccuracy = ((float(len(testXData) - testMistake)) /
len(testXData)) * 100
print '\t Test Accuracy = ', TestAccuracy, '%\n'
print "//End of Question-3: For Grads(Aggressive Perceptron-Batch Mode-Shuffle Report -----------------------------\n\n\n"
def Run_Q3_3_3_mp_batch_noShuffle():
print '***************************************************************************************************'
print "//Question-3.3.2 (Margin Perceptron-batch 3,4,5) No Shuffle Report on Adult Data -----------------------------"
print '***************************************************************************************************'
[XData, YData, FSize, SeenWt] = func.parseInfo(trainFileHandle)
maxVecLen = max(FSize)
for i in range(3, 6):
for j in LRateList:
for k in MarginList:
[Bias, Wvec,
LearningMistake] = func.Perceptron(j, 1, XData, YData,
maxVecLen, i, k, 0, SeenWt)
mistakeCount = func.TestPerceptron(XData, YData, Wvec, Bias,
maxVecLen, k)
TrainAccuracy = (
(float(len(XData) - mistakeCount)) / len(XData)) * 100
print 'Using Learning Rate = ', j, ', number of epochs = ', i, ' Margin = ', k, ', Initialization = random with mean 0 and sd 0.01'
# print '\t Learned Weight Vector = ', Wvec
print '\t Learned Bias = ', Bias
print '\t Mistakes during Learning = ', LearningMistake
print '\t Learning Accuracy = ', TrainAccuracy, '%\n'
if (test_index != -1):
[testXData,
testYData] = func.parseInfoTest(testFilehandle, maxVecLen)
testMistake = func.TestPerceptron(testXData, testYData,
Wvec, Bias, maxVecLen, k)
TestAccuracy = ((float(len(testXData) - testMistake)) /
len(testXData)) * 100
print '\t Test Accuracy = ', TestAccuracy, '%\n'
print "//End of Question-3.3.2(Margin Perceptron-Batch Mode-NoShuffle) Report ------------------------------------\n\n\n"
def Run_Q1():
[XData, YData, FSize] = func.parseInfo(trainFileHandle)
print "YData: ", len(YData), len(XData), FSize
sigmaSq = 1000
lr = 0.1
[WVec, LearningMistake, lr,
neglogdata] = func.LogReg(XData, YData, FSize, sigmaSq, lr, 100, 1)
#[WVec, LearningMistake,lr] = func.LogReg(XData, YData, FSize, sigmaSq, lr, 20,0)
FinalTrainAcc = 100 * float(len(XData) - LearningMistake) / len(XData)
print "Final Training Accuracy = ", FinalTrainAcc, " %"
if (test_index != -1):
[testXData, testYData] = func.parseInfoTest(testFileHandle, FSize)
print "From Test:", len(testXData), len(testYData)
TestMistakes = func.LogRegTest(WVec, testXData, testYData)
FinalTestAcc = 100 * float(len(testXData) -
TestMistakes) / len(testXData)
print "Final Test Accuracy = ", FinalTestAcc, " %"
#print WVec
##------- Negative log-likelihood Plot ---------------
xplot = [neglogdata[i][0] for i in range(0, len(neglogdata))]
yplot = [neglogdata[i][1] for i in range(0, len(neglogdata))]
plt.plot(xplot, yplot, 'bs')
plt.ylabel('Negative Log Likelihood')
plt.xlabel('Epochs')
plt.show()
def Run_Q3_3_1():
print '*********************************************************************************'
print "//Question-3.3.1 Report (Sanity check using Table2 -----------------------------"
print '*********************************************************************************'
[XData, YData, FSize, SeenWt] = func.parseInfo(sanityTable)
maxVecLen = max(FSize)
for i in range(1, 2):
for j in LRateList:
[Bias, Wvec,
LearningMistake] = func.Perceptron(j, 0, XData, YData, maxVecLen,
i, 0, 0, SeenWt)
mistakeCount = func.TestPerceptron(XData, YData, Wvec, Bias,
maxVecLen, 0)
TrainAccuracy = (float(len(XData) - mistakeCount) /
len(XData)) * 100
print 'Using Learning Rate = ', j, ' number of epochs = ', i, 'Initialization = @ default 0 , No-Shuffle'
print '\t Learned Weight Vector = ', Wvec
print '\t Learned Bias = ', Bias
print '\t Mistakes during Learning = ', LearningMistake
print '\t Learning Accuracy = ', TrainAccuracy, '%\n'
print "//End of Question-3.3.1 Report -------------------------------------------------\n\n"
def Run_kvalidate():
[XData, YData, FSize] = func.parseInfo(trainFileHandle)
bestSigmaSq = 0
bestlr = 0
MaxAccuracy = 0
epochs = 20
#lr = 0.00001
kfoldData = []
print "\n################################################################"
print " Starting 5 Fold Cross Validation "
print "################################################################\n"
for lr0 in lrlist:
for sigmaSq in SigmaSqList:
blockSize = len(XData) / foldValue
testAcc = 0
trainAcc = 0
for k in range(0, foldValue):
#print "SigmaSq = ",sigmaSq, ", LR0 = ",lr0
KXTest = XData[k * blockSize:(k + 1) * blockSize]
KYTest = YData[k * blockSize:(k + 1) * blockSize]
KXData = [
XData[i] for i in range(0, len(XData))
if (i < k * blockSize or i >= (k + 1) * blockSize)
]
KYData = [
YData[i] for i in range(0, len(XData))
if (i < k * blockSize or i >= (k + 1) * blockSize)
]
[Wvec, trainMist, lr] = func.LogReg(KXData, KYData, FSize,
sigmaSq, lr0, epochs, 0)
testMist = func.LogRegTest(Wvec, KXTest, KYTest)
trainAcc = trainAcc + 100 * float(
(len(KXData) - trainMist)) / len(KXData)
testAcc = testAcc + 100 * float(
(len(KXTest) - testMist)) / len(KXTest)
avgtrainacc = trainAcc / foldValue
avgtestacc = testAcc / foldValue
print "HyperParameter INFO:: %15s" % "SigmaSquare", " = %6s" % sigmaSq, ": %20s" % " LearningRate", " = %6s" % lr0, ": %10s" % " TrainAcc", " = %8s" % avgtrainacc, ": %10s" % " TestAcc", " = %8s " % avgtestacc
kfoldData.append([sigmaSq, lr0, avgtrainacc, avgtestacc])
if (MaxAccuracy < avgtestacc):
MaxAccuracy = avgtestacc
bestSigmaSq = sigmaSq
bestlr = lr0
print "HyperParameter INFO:: %15s" % "Best SigmaSq", " = %6s" % bestSigmaSq, ": %20s" % "BestLearningRate", " = %6s" % bestlr, "\n\n"
print "\n\n******************************************************"
print "HyperParameter Info:: Best Found SigmaSquare = %8s" % bestSigmaSq, ": Best Initial Learning Rate = %8s" % bestlr
print "******************************************************\n\n"
##----------- Cross Validation Report ------------------------
# for i in kfoldData:
# print "SigmaSquare = ", i[0], "LearningRate = ", i[1], ", TrainAcc = ",i[2], ", TestAcc = ",i[3]
##----------- Now learn on the entire data set --------------------##
print "Starting Final Training (On 80 Epochs)..................."
[Wvec, trainMist, lr, neglogdata] = func.LogReg(XData, YData, FSize,
bestSigmaSq, bestlr, 80, 1)
print "Training Completed ......................."
FinalTrainAcc = 100 * float(len(XData) - trainMist) / len(XData)
print "Final Training Accuracy = %8s" % FinalTrainAcc, "%\n\n"
#print "Train Data size = ", len(XData)
if (test_index != -1):
print "Initiating Final Testing ................."
[testXData, testYData] = func.parseInfoTest(testFileHandle, FSize)
TestMistakes = func.LogRegTest(Wvec, testXData, testYData)
#print "Test Data size = ", len(testXData)
print "Final Testing Completed................."
FinalTestAcc = 100 * float(len(testXData) -
TestMistakes) / len(testXData)
print "Final Test Accuracy = %8s" % FinalTestAcc, "%\n\n"
##------- Negative log-likelihood Plot ---------------
print "\n\n###########################################"
print "Plotting the Negative Log Likelihood"
print "###########################################"
xplot = [neglogdata[i][0] for i in range(0, len(neglogdata))]
yplot = [neglogdata[i][1] for i in range(0, len(neglogdata))]
plt.plot(xplot, yplot)
plt.ylabel('Negative Log Likelihood')
plt.xlabel('Epochs')
plt.show()
w_temp2 = {}
TrainToolsDict_temp = {}
accuracy_lst = []
##--------- Get the training data ---------##
dataFileName = Dpath + "/Training/" + tool + ".data"
allData = open(dataFileName, "rb").read().splitlines()
for t in range(epoch):
random.shuffle(allData)
trainLength = int(len(allData) * 0.6)
testLength = len(allData) - trainLength
trainData = []
for i in range(trainLength):
trainData.append(random.choice(allData))
[xTrainData, yTrainData] = func.parseInfo(trainData)
w_temp2[t] = linearRegression.LinearRegression(
xTrainData, yTrainData)
##---- Get the error norm and score on the training data -----##
TrainToolsDict_temp[t] = func.getPredictedScoreError(
xTrainData, yTrainData, w_temp2[t])
testData = [item for item in allData if item not in trainData]
[xTestData, yTestData] = func.parseInfo(testData)
accuracy_lst.append(func.accuracy(w_temp2[t], xTestData,
yTestData))
#func.reportAccuracy(tool, accuracy_lin[tool])
## --- Get the mean of acuuracies --- ##
accuracy_lin[tool] = sum(accuracy_lst) / len(accuracy_lst) * 100
TrainToolsDict = {}
for tool in trainTools:
##--------- Get the training data ---------##
trainFileName = Dpath+"/Training/"+tool+".data"
#trainData = csv.reader(open(trainFileName, "rb"), delimiter=" ")
trainData = open(trainFileName, "rb").read().splitlines()
trainLength = int(len(trainData) * 0.6)
testLength = len(trainData) - trainLength
trainData2 = []
for i in range(trainLength):
trainData2.append(random.choice(trainData))
[xdata,ydata] = func.parseInfo(trainData2)
w_temp = linearRegression.LinearRegression(xdata, ydata)
w[tool] = w_temp
##---- Get the error norm and score on the training data -----##
TrainToolsDict[tool] = func.getPredictedScoreError(xdata,ydata, w_temp)
#print "Completed Linear Regression learning .....................\n"
func.reportTrend(TrainToolsDict, "Linear: Training on SVComp14")
# print "w:" , w
##------ Testing on SVComp15 (Predict the overall Winner)----------
if(testId!=-1 and trainId!=-1): ##indicates training dataStructure is available
TestToolDict = {}