def readDigitData(trainingSize=100, testSize=100):
rootdata = "digitdata/"
# loading digits data
rawTrainingData = samples.loadDataFile(
rootdata + "trainingimages", trainingSize, DIGIT_DATUM_WIDTH, DIGIT_DATUM_HEIGHT
)
trainingLabels = samples.loadLabelsFile(rootdata + "traininglabels", trainingSize)
rawValidationData = samples.loadDataFile(
rootdata + "validationimages", TEST_SET_SIZE, DIGIT_DATUM_WIDTH, DIGIT_DATUM_HEIGHT
)
validationLabels = samples.loadLabelsFile(rootdata + "validationlabels", TEST_SET_SIZE)
rawTestData = samples.loadDataFile("digitdata/testimages", testSize, DIGIT_DATUM_WIDTH, DIGIT_DATUM_HEIGHT)
testLabels = samples.loadLabelsFile("digitdata/testlabels", testSize)
try:
print "Extracting features..."
featureFunction = dataClassifier.basicFeatureExtractorDigit
trainingData = map(featureFunction, rawTrainingData)
validationData = map(featureFunction, rawValidationData)
testData = map(featureFunction, rawTestData)
except:
display("An exception was raised while extracting basic features: \n %s" % getExceptionTraceBack())
return (
trainingData,
trainingLabels,
validationData,
validationLabels,
rawTrainingData,
rawValidationData,
testData,
testLabels,
rawTestData,
)
def get_neuron_test_data():
test_data = samples.loadDataFile("digitdata/testimages", 1000, 28,28)
test_labels = np.array(samples.loadLabelsFile("digitdata/testlabels", 1000))
test_labels = test_labels == 3
featurized_test_data = np.array(map(dcu.simple_image_featurization, test_data))
return test_data, featurized_test_data, test_labels
def get_neuron_training_data():
training_data = samples.loadDataFile("digitdata/trainingimages", num_train_examples, 28, 28)
training_labels = np.array(samples.loadLabelsFile("digitdata/traininglabels", num_train_examples))
training_labels = training_labels == 3
featurized_training_data = np.array(map(dcu.simple_image_featurization, training_data))
return training_data, featurized_training_data, training_labels
def runClassifier(args, options):
classifier = args['classifier']
printImage = args['printImage']
# Load data
numTraining = options.training
numTest = options.test
if(options.data=="faces"):
print "loading face data set"
rawTrainingData = samples.loadDataFile("facedata/facedatatrain",FACE_DATUM_WIDTH,FACE_DATUM_HEIGHT)
trainingLabels = samples.loadLabelsFile("facedata/facedatatrainlabels")
rawValidationData = samples.loadDataFile("facedata/facedatavalidation",FACE_DATUM_WIDTH,FACE_DATUM_HEIGHT)
validationLabels = samples.loadLabelsFile("facedata/facedatavalidationlabels")
rawTestData = samples.loadDataFile("facedata/facedatatest", FACE_DATUM_WIDTH,FACE_DATUM_HEIGHT)
testLabels = samples.loadLabelsFile("facedata/facedatatestlabels")
rawTrainingData,trainingLabels=randomSample(rawTrainingData,trainingLabels,numTraining)
rawTestData,testLabels=randomSample(rawTestData,testLabels,numTest)
else:
print "loading digit data set"
rawTrainingData = samples.loadDataFile("digitdata/trainingimages",DIGIT_DATUM_WIDTH,DIGIT_DATUM_HEIGHT)
trainingLabels = samples.loadLabelsFile("digitdata/traininglabels")
rawValidationData = samples.loadDataFile("digitdata/validationimages",DIGIT_DATUM_WIDTH,DIGIT_DATUM_HEIGHT)
validationLabels = samples.loadLabelsFile("digitdata/validationlabels")
rawTestData = samples.loadDataFile("digitdata/testimages",DIGIT_DATUM_WIDTH,DIGIT_DATUM_HEIGHT)
testLabels = samples.loadLabelsFile("digitdata/testlabels")
rawTrainingData, trainingLabels = randomSample(rawTrainingData, trainingLabels, numTraining)
rawTestData, testLabels = randomSample(rawTestData, testLabels, numTest)
print "Extracting features..."
if (options.classifier == "linear_svm"):
if (options.data == "faces"):
featureFunction = HogFeatureFaceImg
else:
featureFunction=HogFeatureImgDigit
trainingData = map(featureFunction, rawTrainingData)
trainingData=np.array(trainingData).transpose()
validationData=map(featureFunction, rawValidationData)
validationData = np.array(validationData).transpose()
testData=map(featureFunction, rawTestData)
testData = np.array(testData).transpose()
else:
if (options.data == "faces"):
featureFunction = enhancedFeatureExtractorFace
else:
featureFunction = enhancedFeatureExtractorDigit
trainingData = map(featureFunction, rawTrainingData)
validationData = map(featureFunction, rawValidationData)
testData = map(featureFunction, rawTestData)
print "Training..."
start = timeit.default_timer()
classifier.train(trainingData, trainingLabels, validationData, validationLabels)
stop = timeit.default_timer()
print stop - start, " s"
print "Validating..."
guesses = classifier.classify(validationData)
correct = [guesses[i] == validationLabels[i] for i in range(len(validationLabels))].count(True)
print str(correct), ("correct out of " + str(len(validationLabels)) + " (%.1f%%).") % (100.0 * correct / len(validationLabels))
print "Testing..."
guesses = classifier.classify(testData)
correct = [guesses[i] == testLabels[i] for i in range(len(testLabels))].count(True)
print str(correct), ("correct out of " + str(len(testLabels)) + " (%.1f%%).") % (100.0 * correct / len(testLabels))
analysis(classifier, guesses, testLabels, testData, rawTestData, printImage)
def runClassifier(args, options):
featureFunction = args['featureFunction']
classifier = args['classifier']
printImage = args['printImage']
# Load data
numTraining = options.training
if(options.data=="faces"):
rawTrainingData = samples.loadDataFile("facedata/facedatatrain", numTraining,FACE_DATUM_WIDTH,FACE_DATUM_HEIGHT)
trainingLabels = samples.loadLabelsFile("facedata/facedatatrainlabels", numTraining)
rawValidationData = samples.loadDataFile("facedata/facedatatrain", TEST_SET_SIZE,FACE_DATUM_WIDTH,FACE_DATUM_HEIGHT)
validationLabels = samples.loadLabelsFile("facedata/facedatatrainlabels", TEST_SET_SIZE)
rawTestData = samples.loadDataFile("facedata/facedatatest", TEST_SET_SIZE,FACE_DATUM_WIDTH,FACE_DATUM_HEIGHT)
testLabels = samples.loadLabelsFile("facedata/facedatatestlabels", TEST_SET_SIZE)
else:
rawTrainingData = samples.loadDataFile("digitdata/trainingimages", numTraining,DIGIT_DATUM_WIDTH,DIGIT_DATUM_HEIGHT)
trainingLabels = samples.loadLabelsFile("digitdata/traininglabels", numTraining)
rawValidationData = samples.loadDataFile("digitdata/validationimages", TEST_SET_SIZE,DIGIT_DATUM_WIDTH,DIGIT_DATUM_HEIGHT)
validationLabels = samples.loadLabelsFile("digitdata/validationlabels", TEST_SET_SIZE)
rawTestData = samples.loadDataFile("digitdata/testimages", TEST_SET_SIZE,DIGIT_DATUM_WIDTH,DIGIT_DATUM_HEIGHT)
testLabels = samples.loadLabelsFile("digitdata/testlabels", TEST_SET_SIZE)
# Extract features
print "Extracting features..."
trainingData = map(featureFunction, rawTrainingData)
validationData = map(featureFunction, rawValidationData)
testData = map(featureFunction, rawTestData)
# Conduct training and testing
print "Training..."
classifier.train(trainingData, trainingLabels, validationData, validationLabels)
print "Validating..."
guesses = classifier.classify(validationData)
correct = [guesses[i] == validationLabels[i] for i in range(len(validationLabels))].count(True)
print str(correct), ("correct out of " + str(len(validationLabels)) + " (%.1f%%).") % (100.0 * correct / len(validationLabels))
print "Testing..."
guesses = classifier.classify(testData)
correct = [guesses[i] == testLabels[i] for i in range(len(testLabels))].count(True)
print str(correct), ("correct out of " + str(len(testLabels)) + " (%.1f%%).") % (100.0 * correct / len(testLabels))
analysis(classifier, guesses, testLabels, testData, rawTestData, printImage)
# do odds ratio computation if specified at command line
if((options.odds) & (options.classifier != "mostFrequent")):
label1, label2 = options.label1, options.label2
features_odds = classifier.findHighOddsFeatures(label1,label2)
if(options.classifier == "naiveBayes" or options.classifier == "nb"):
string3 = "=== Features with highest odd ratio of label %d over label %d ===" % (label1, label2)
else:
string3 = "=== Features for which weight(label %d)-weight(label %d) is biggest ===" % (label1, label2)
print string3
printImage(features_odds)
def runClassifier(args, options):
featureFunction = args['featureFunction']
classifier = args['classifier']
printImage = args['printImage']
# Load data
numTraining = options.training
if(options.data=="faces"):
rawTrainingData = samples.loadDataFile("facedata/facedatatrain", numTraining,FACE_DATUM_WIDTH,FACE_DATUM_HEIGHT)
trainingLabels = samples.loadLabelsFile("facedata/facedatatrainlabels", numTraining)
rawValidationData = samples.loadDataFile("facedata/facedatatrain", TEST_SET_SIZE,FACE_DATUM_WIDTH,FACE_DATUM_HEIGHT)
validationLabels = samples.loadLabelsFile("facedata/facedatatrainlabels", TEST_SET_SIZE)
rawTestData = samples.loadDataFile("facedata/facedatatest", TEST_SET_SIZE,FACE_DATUM_WIDTH,FACE_DATUM_HEIGHT)
testLabels = samples.loadLabelsFile("facedata/facedatatestlabels", TEST_SET_SIZE)
else:
rawTrainingData = samples.loadDataFile("digitdata/trainingimages", numTraining,DIGIT_DATUM_WIDTH,DIGIT_DATUM_HEIGHT)
trainingLabels = samples.loadLabelsFile("digitdata/traininglabels", numTraining)
rawValidationData = samples.loadDataFile("digitdata/validationimages", TEST_SET_SIZE,DIGIT_DATUM_WIDTH,DIGIT_DATUM_HEIGHT)
validationLabels = samples.loadLabelsFile("digitdata/validationlabels", TEST_SET_SIZE)
rawTestData = samples.loadDataFile("digitdata/testimages", TEST_SET_SIZE,DIGIT_DATUM_WIDTH,DIGIT_DATUM_HEIGHT)
testLabels = samples.loadLabelsFile("digitdata/testlabels", TEST_SET_SIZE)
# Extract features
print "Extracting features..."
trainingData = map(featureFunction, rawTrainingData)
validationData = map(featureFunction, rawValidationData)
testData = map(featureFunction, rawTestData)
# Conduct training and testing
print "Training..."
classifier.train(trainingData, trainingLabels, validationData, validationLabels)
print "Validating..."
guesses = classifier.classify(validationData)
correct = [guesses[i] == validationLabels[i] for i in range(len(validationLabels))].count(True)
print str(correct), ("correct out of " + str(len(validationLabels)) + " (%.1f%%).") % (100.0 * correct / len(validationLabels))
print "Testing..."
guesses = classifier.classify(testData)
correct = [guesses[i] == testLabels[i] for i in range(len(testLabels))].count(True)
print str(correct), ("correct out of " + str(len(testLabels)) + " (%.1f%%).") % (100.0 * correct / len(testLabels))
analysis(classifier, guesses, testLabels, testData, rawTestData, printImage)
def run_digits_n_times(digit, iterations, sample_percentage, trainingpath,
labelspath):
all_final_weights = []
all_final_accuracies = []
images = []
labels = []
'''
Load either digit training data or face training data
'''
n_images == 5000
images = samples.loadDataFile(trainingpath, n_images, 28, 28)
labels = samples.loadLabelsFile(labelspath, n_images)
'''
Wrap datum in Node object, store Node objects in a list, list index = datum/node's label
'''
'''
Compute weights / run perceptron n times, where n = iterations parameter
'''
for a in range(iterations):
# Functions_list = [pf.avg_horizontal_line_length, pf.variance_horizontal_line_length]
images_sample, labels_sample, visited = sample_digits(
digit, sample_percentage, images, labels)
'''
!!!! SUBJECT TO CHANGE !!!
'''
# Featureslist = compute_features(functions_list, images_sample, typeflag)
featureslist = compute_features2(images_sample)
# Weights = initialize_weights(len(functions_list), 0)
weights = initialize_weights(28 * 28, 0)
'''
Run Perceptron
'''
start = time.time()
final_weights = compute_weights(weights, featureslist, labels_sample)
elapsed = time.time() - start
'''
Validate weights
'''
# Why was the first parameter 1???????? should be digit?? re run digits with digit instead of 1
accuracy = validate_weights(digit, final_weights)
###
print(
str(digit) + ': ' + str(elapsed) + ' ~~ ' + 'sample percent: ' +
str(sample_percentage) + ' ~~ ' + str(accuracy) + '%')
basepath = './TrainingDigitsResults120/TrainingDigitsResults' + str(
digit) + '/' + str(sample_percentage) + '_percent.txt'
with open(basepath, 'w') as file:
for weight in range(len(weights)):
if weight == len(weights) - 1:
file.write(str(weights[weight]) + '\n')
else:
file.write(str(weights[weight]) + ' ')
file.write(
str(round(elapsed, 2)) + 's' + ' ' +
str(round(float(accuracy) / float(100), 2)))
'''
Record computed weights and accuracy for this training iteration
'''
all_final_weights.append(final_weights)
all_final_accuracies.append(accuracy)
'''
Record mean accuracy for all iterations
'''
'''
def validate_digits120():
start = time.time()
weights_vectors = []
base_path = './TrainingDigitsResults120/TrainingDigitsResults'
for i in range(0, 10):
load_path = base_path + str(i) + '/100_percent.txt'
with open(load_path, 'r') as file:
lines = file.readlines()
weightsstrings = lines[0].split()
weights = [float(weight) for weight in weightsstrings]
print(weights)
weights_vectors.append(weights)
file.close()
images = samples.loadDataFile('digitdata/validationimages', 1000, 28, 28)
labels = samples.loadLabelsFile('digitdata/validationlabels', 1000)
featureslist = compute_features2(images)
'''
For each image, apply all weight vectors to image
Keep track of each sum in a sums list for the current image,
Choose the sum that:
The highest value
The sum's index in the sums dictionary is the same as the weight vector's index
The weight vector's index is the same as the digit's designated/chosen weights
This index will be what digit you are guessing the image to be
Compare this index/guessed number with the labels[image] value, if they are the same, append True to results
if false, append False to results
'''
results = []
for image in range(len(images)):
sums = []
for weightsvector in range(len(weights_vectors)):
sum = float(0)
for weight in range(len(weights_vectors[weightsvector])):
sum += (weights_vectors[weightsvector][weight] *
featureslist[image][weight])
# Add the sum to the sums dictionary
sums.append(sum)
max = float('-inf')
index = -1
for z in range(10):
if sums[z] > max:
max = sums[z]
index = z
# print(index, end=" --> ")
# print(labels[image])
# Keep track of all guesses vs. labels in a tuples list (guess, label)
results.append((index, labels[image]))
correctcount = float(0)
for t in results:
if t[0] == t[1]:
correctcount += float(1)
print('Digits accuracy: ' +
str(round((float(correctcount) * 100 / float(len(labels))), 1)) +
'%')
def runClassifier(args, options):
featureFunction = args['featureFunction']
classifier = args['classifier']
# Load data
dataset = options.dataset
numTraining = options.training
numTest = options.test
if dataset == 'd1':
rawTrainingData = samples.loadDataFile("data/D1/training_data",
numTraining)
trainingLabels = samples.loadLabelsFile("data/D1/training_labels",
numTraining)
rawTestData = samples.loadDataFile("data/D1/test_data", numTest)
testLabels = samples.loadLabelsFile("data/D1/test_labels", numTest)
else:
rawTrainingData = samples.loadDataFile("data/D2/training_data",
numTraining)
trainingLabels = samples.loadLabelsFile("data/D2/training_labels",
numTraining)
rawTestData = samples.loadDataFile("data/D2/test_data", numTest)
testLabels = samples.loadLabelsFile("data/D2/test_labels", numTest)
# Extract features
print "Extracting features..."
trainingData = map(featureFunction, rawTrainingData)
testData = map(featureFunction, rawTestData)
# Conduct training and testing
print "Training..."
classifier.train(trainingData, trainingLabels, testData, testLabels,
options.validate)
guesses = classifier.classify(trainingData)
correct = [
guesses[i] == trainingLabels[i] for i in range(len(trainingLabels))
].count(True)
if (options.classifier == "1vr"):
f = open("perceptron1vr_train.csv", "a")
f.write(
str(len(trainingData)) + "," +
str(100 * correct / (1.0 * (len(trainingData)))) + '\n')
f.close()
print "Testing..."
guesses = classifier.classify(testData)
# for i in range(len(testLabels)):
# if guesses[i] != testLabels[i]:
# print trainingData[i], guesses[i], testLabels[i]
correct = [guesses[i] == testLabels[i]
for i in range(len(testLabels))].count(True)
print str(correct), ("correct out of " + str(len(testLabels)) +
" (%.1f%%).") % (100.0 * correct / len(testLabels))
if (options.classifier == "1vr"):
f = open("perceptron1vr_test.csv", "a")
f.write(
str(len(trainingData)) + "," + str(100 * correct /
(1.0 * (len(testData)))) + '\n')
f.close()
import minicontest
import samples
import sys
import util
import pickle
from dataClassifier import DIGIT_DATUM_HEIGHT, DIGIT_DATUM_WIDTH, contestFeatureExtractorDigit
TEST_SIZE = 1000
MINICONTEST_PATH = "minicontest_output.pickle"
if __name__ == '__main__':
print "Loading training data"
rawTrainingData = samples.loadDataFile("digitdata/trainingimages", 5000,
DIGIT_DATUM_WIDTH,
DIGIT_DATUM_HEIGHT)
trainingLabels = samples.loadLabelsFile("digitdata/traininglabels", 5000)
rawValidationData = samples.loadDataFile("digitdata/validationimages", 100,
DIGIT_DATUM_WIDTH,
DIGIT_DATUM_HEIGHT)
validationLabels = samples.loadLabelsFile("digitdata/validationlabels",
100)
rawTestData = samples.loadDataFile("digitdata/testimages", TEST_SIZE,
DIGIT_DATUM_WIDTH, DIGIT_DATUM_HEIGHT)
featureFunction = contestFeatureExtractorDigit
legalLabels = range(10)
classifier = minicontest.contestClassifier(legalLabels)
print "Extracting features..."
def runClassifier(args, options):
featureFunction = args['featureFunction']
classifier = args['classifier']
printImage = args['printImage']
# Load data
numTraining = options.training
numTest = options.test
if (options.data == "faces"):
rawTrainingData = samples.loadDataFile("facedata/facedatatrain",
numTraining, FACE_DATUM_WIDTH,
FACE_DATUM_HEIGHT)
trainingLabels = samples.loadLabelsFile("facedata/facedatatrainlabels",
numTraining)
rawValidationData = samples.loadDataFile("facedata/facedatatrain",
numTest, FACE_DATUM_WIDTH,
FACE_DATUM_HEIGHT)
validationLabels = samples.loadLabelsFile(
"facedata/facedatatrainlabels", numTest)
rawTestData = samples.loadDataFile("facedata/facedatatest", numTest,
FACE_DATUM_WIDTH, FACE_DATUM_HEIGHT)
testLabels = samples.loadLabelsFile("facedata/facedatatestlabels",
numTest)
else:
rawTrainingData = samples.loadDataFile("digitdata/trainingimages",
numTraining, DIGIT_DATUM_WIDTH,
DIGIT_DATUM_HEIGHT)
trainingLabels = samples.loadLabelsFile("digitdata/traininglabels",
numTraining)
rawValidationData = samples.loadDataFile("digitdata/validationimages",
numTest, DIGIT_DATUM_WIDTH,
DIGIT_DATUM_HEIGHT)
validationLabels = samples.loadLabelsFile("digitdata/validationlabels",
numTest)
rawTestData = samples.loadDataFile("digitdata/testimages", numTest,
DIGIT_DATUM_WIDTH,
DIGIT_DATUM_HEIGHT)
testLabels = samples.loadLabelsFile("digitdata/testlabels", numTest)
# Extract features
print("Extracting features...")
#creates the features for the training
trainingData = map(featureFunction, rawTrainingData)
#creates the features for validation
validationData = map(featureFunction, rawValidationData)
#features for the testdata
testData = map(featureFunction, rawTestData)
# Conduct training and testing
copydata = copy(trainingData)
copylabel = copy(trainingLabels)
begin = time.time()
for percentageOfData in range(10):
start = time.time()
trainingData = copy(copydata)
trainingLabels = copy(copylabel)
del trainingData[numTraining / 10 * (percentageOfData + 1):numTraining]
del trainingLabels[numTraining / 10 *
(percentageOfData + 1):numTraining]
print("Training: %d " % (numTraining / 10 * (percentageOfData + 1)))
# basically trains the data
classifier.train(trainingData, trainingLabels, validationData,
validationLabels)
print("Validating...")
guesses = classifier.classify(validationData)
correct = [
guesses[i] == validationLabels[i]
for i in range(len(validationLabels))
].count(True)
print(str(correct),
("correct out of " + str(len(validationLabels)) + " (%.1f%%).") %
(100.0 * correct / len(validationLabels)))
print("Testing...")
guesses = classifier.classify(testData)
correct = [
guesses[i] == testLabels[i] for i in range(len(testLabels))
].count(True)
print(str(correct),
("correct out of " + str(len(testLabels)) + " (%.1f%%).") %
(100.0 * correct / len(testLabels)))
elapsed_time_fl = (time.time() - start)
print("%d secs" % (elapsed_time_fl))
totaltime = (time.time() - begin)
print("Total time: %d" % (totaltime))
#analysis(classifier, guesses, testLabels, testData, rawTestData, printImage)
# do odds ratio computation if specified at command line
if ((options.odds) & (options.classifier == "naiveBayes" or
(options.classifier == "nb"))):
label1, label2 = options.label1, options.label2
features_odds = classifier.findHighOddsFeatures(label1, label2)
if (options.classifier == "naiveBayes" or options.classifier == "nb"):
string3 = "=== Features with highest odd ratio of label %d over label %d ===" % (
label1, label2)
else:
string3 = "=== Features for which weight(label %d)-weight(label %d) is biggest ===" % (
label1, label2)
print(string3)
printImage(features_odds)
if ((options.weights) & (options.classifier == "perceptron")):
for l in classifier.legalLabels:
features_weights = classifier.findHighWeightFeatures(l)
print("=== Features with high weight for label %d ===" % l)
printImage(features_weights)
import numpy as np
def writeLabeledData(prefix, labeled_data):
datums, labels = zip(*labeled_data)
with open(prefix + "images", 'w') as f:
for datum in datums:
f.write(str(datum) + "\n")
f.close()
with open(prefix + "labels", 'w') as f:
for label in labels:
f.write(str(label) + "\n")
f.close()
rawTrainingData = samples.loadDataFile("digitdata/trainingimages", 5000,DIGIT_DATUM_WIDTH,DIGIT_DATUM_HEIGHT)
trainingLabels = samples.loadLabelsFile("digitdata/traininglabels", 5000)
rawValidationData = samples.loadDataFile("digitdata/validationimages", 1000,DIGIT_DATUM_WIDTH,DIGIT_DATUM_HEIGHT)
validationLabels = samples.loadLabelsFile("digitdata/validationlabels", 1000)
rawTestData = samples.loadDataFile("digitdata/testimages", 1000,DIGIT_DATUM_WIDTH,DIGIT_DATUM_HEIGHT)
testLabels = samples.loadLabelsFile("digitdata/testlabels", 1000)
all_data = rawTrainingData + rawValidationData + rawTestData
all_labels = trainingLabels + validationLabels + testLabels
labeled_data = zip(all_data, all_labels)
perm = np.random.permutation(len(labeled_data))
permuted_data = []
def runClassifier(args, options):
featureFunction = args['featureFunction']
classifier = args['classifier']
printImage = args['printImage']
# Load data
numTraining = options.training
numTest = options.test
rawTrainingData = samples.loadDataFile("data/digitdata/trainingimages",
numTraining, DIGIT_DATUM_WIDTH,
DIGIT_DATUM_HEIGHT)
trainingLabels = samples.loadLabelsFile("data/digitdata/traininglabels",
numTraining)
completeRawTrainingData = samples.loadDataFile(
"data/digitdata/trainingimages", 5000, DIGIT_DATUM_WIDTH,
DIGIT_DATUM_HEIGHT)
completeTrainingLabels = samples.loadLabelsFile(
"data/digitdata/traininglabels", 5000)
rawValidationData = samples.loadDataFile("data/digitdata/validationimages",
numTest, DIGIT_DATUM_WIDTH,
DIGIT_DATUM_HEIGHT)
validationLabels = samples.loadLabelsFile(
"data/digitdata/validationlabels", numTest)
rawTestData = samples.loadDataFile("data/digitdata/testimages", numTest,
DIGIT_DATUM_WIDTH, DIGIT_DATUM_HEIGHT)
testLabels = samples.loadLabelsFile("data/digitdata/testlabels", numTest)
# Extract features
print "Extracting features..."
trainingData = map(featureFunction, rawTrainingData)
completeTrainingData = map(featureFunction, completeRawTrainingData)
validationData = map(featureFunction, rawValidationData)
testData = map(featureFunction, rawTestData)
# Conduct training and testing
print "Training..."
classifier.train(trainingData, trainingLabels, validationData,
validationLabels)
print "Validating..."
guesses = classifier.classify(validationData)
correct = [
guesses[i] == validationLabels[i] for i in range(len(validationLabels))
].count(True)
print str(correct), ("correct out of " + str(len(validationLabels)) +
" (%.1f%%).") % (100.0 * correct /
len(validationLabels))
print "Testing..."
guesses = classifier.classify(testData)
correct = [guesses[i] == testLabels[i]
for i in range(len(testLabels))].count(True)
print str(correct), ("correct out of " + str(len(testLabels)) +
" (%.1f%%).") % (100.0 * correct / len(testLabels))
print "Testing training data..."
guesses = classifier.classify(completeTrainingData)
correct = [
guesses[i] == completeTrainingLabels[i]
for i in range(len(completeTrainingLabels))
].count(True)
print str(correct), ("correct out of " + str(len(completeTrainingLabels)) +
" (%.1f%%).") % (100.0 * correct /
len(completeTrainingLabels))
analysis(classifier, guesses, testLabels, testData, rawTestData,
printImage)
if ((options.classifier == "perceptron")):
for l in classifier.legalLabels:
features_weights = classifier.findHighWeightFeatures(l)
print("=== Features with high weight for label %d ===" % l)
printImage(features_weights)
def runClassifier(args, options):
featureFunction = args['featureFunction']
classifier = args['classifier']
printImage = args['printImage']
# Load data
numTraining = options.training
numTest = options.test
if (options.data == "faces"):
rawTrainingData = samples.loadDataFile("facedata/facedatatrain",
numTraining, FACE_DATUM_WIDTH,
FACE_DATUM_HEIGHT)
trainingLabels = samples.loadLabelsFile("facedata/facedatatrainlabels",
numTraining)
rawValidationData = samples.loadDataFile("facedata/facedatatrain",
numTest, FACE_DATUM_WIDTH,
FACE_DATUM_HEIGHT)
validationLabels = samples.loadLabelsFile(
"facedata/facedatatrainlabels", numTest)
rawTestData = samples.loadDataFile("facedata/facedatatest", numTest,
FACE_DATUM_WIDTH, FACE_DATUM_HEIGHT)
testLabels = samples.loadLabelsFile("facedata/facedatatestlabels",
numTest)
else:
rawTrainingData = samples.loadDataFile("digitdata/trainingimages",
numTraining, DIGIT_DATUM_WIDTH,
DIGIT_DATUM_HEIGHT)
trainingLabels = samples.loadLabelsFile("digitdata/traininglabels",
numTraining)
rawValidationData = samples.loadDataFile("digitdata/trainingimages",
numTest, DIGIT_DATUM_WIDTH,
DIGIT_DATUM_HEIGHT)
validationLabels = samples.loadLabelsFile("digitdata/traininglabels",
numTest)
rawTestData = samples.loadDataFile("digitdata/testimages", numTest,
DIGIT_DATUM_WIDTH,
DIGIT_DATUM_HEIGHT)
testLabels = samples.loadLabelsFile("digitdata/testlabels", numTest)
# Extract features
print("Extracting features...")
""" old
trainingData = map(featureFunction, rawTrainingData)
validationData = map(featureFunction, rawValidationData)
testData = map(featureFunction, rawTestData)
"""
trainingData = []
validationData = []
testData = []
for d in rawTrainingData:
trainingData.append(featureFunction(d))
for d in rawValidationData:
validationData.append(featureFunction(d))
for d in rawTestData:
testData.append(featureFunction(d))
# Conduct training and testing
print("Training...")
trainTimeStart = time.time()
classifier.train(trainingData, trainingLabels, validationData,
validationLabels)
print("Training completed in %s seconds." % (time.time() - trainTimeStart))
print("Validating...")
guesses = classifier.classify(validationData)
correct = [
guesses[i] == validationLabels[i] for i in range(len(validationLabels))
].count(True)
print(str(correct),
("correct out of " + str(len(validationLabels)) + " (%.1f%%).") %
(100.0 * correct / len(validationLabels)))
print("Testing...")
guesses = classifier.classify(testData)
correct = [guesses[i] == testLabels[i]
for i in range(len(testLabels))].count(True)
print(str(correct),
("correct out of " + str(len(testLabels)) + " (%.1f%%).") %
(100.0 * correct / len(testLabels)))
analysis(classifier, guesses, testLabels, testData, rawTestData,
printImage)
# do odds ratio computation if specified at command line
if ((options.odds) & (options.classifier == "naiveBayes" or
(options.classifier == "nb"))):
label1, label2 = options.label1, options.label2
features_odds = classifier.findHighOddsFeatures(label1, label2)
if (options.classifier == "naiveBayes" or options.classifier == "nb"):
string3 = "=== Features with highest odd ratio of label %d over label %d ===" % (
label1, label2)
else:
string3 = "=== Features for which weight(label %d)-weight(label %d) is biggest ===" % (
label1, label2)
print(string3)
printImage(features_odds)
if ((options.weights) & (options.classifier == "perceptron")):
for l in classifier.legalLabels:
features_weights = classifier.findHighWeightFeatures(l)
print("=== Features with high weight for label %d ===" % l)
printImage(features_weights)
def runClassifier(args, options):
featureFunction = args['featureFunction']
classifier = args['classifier']
printImage = args['printImage']
# Load data
# numTraining = options.training
# numTest = options.test
if options.specialMode:
numberOfTestPoints = 150 if options.data == "faces" else 1000
numberOfValidationPoints = 301 if options.data == "faces" else 1000
totalTrainData = 451 if options.data == "faces" else 5000
numValidation = numberOfValidationPoints
numTest = numberOfTestPoints
numTraining = totalTrainData
# Load Test Data Set
if options.data == "faces":
rawTestData = samples.loadDataFile("facedata/facedatatest",
numTest, FACE_DATUM_WIDTH,
FACE_DATUM_HEIGHT)
testLabels = samples.loadLabelsFile("facedata/facedatatestlabels",
numTest)
rawValidationData = samples.loadDataFile("facedata/facedatatrain",
numValidation,
FACE_DATUM_WIDTH,
FACE_DATUM_HEIGHT)
validationLabels = samples.loadLabelsFile(
"facedata/facedatatrainlabels", numValidation)
rawTrainingData = samples.loadDataFile("facedata/facedatatrain",
numTraining,
FACE_DATUM_WIDTH,
FACE_DATUM_HEIGHT)
trainingLabels = samples.loadLabelsFile(
"facedata/facedatatrainlabels", numTraining)
else:
rawTestData = samples.loadDataFile("digitdata/testimages", numTest,
DIGIT_DATUM_WIDTH,
DIGIT_DATUM_HEIGHT)
testLabels = samples.loadLabelsFile("digitdata/testlabels",
numTest)
rawValidationData = samples.loadDataFile(
"digitdata/validationimages", numValidation, DIGIT_DATUM_WIDTH,
DIGIT_DATUM_HEIGHT)
validationLabels = samples.loadLabelsFile(
"digitdata/validationlabels", numValidation)
rawTrainingData = samples.loadDataFile("digitdata/trainingimages",
numTraining,
DIGIT_DATUM_WIDTH,
DIGIT_DATUM_HEIGHT)
trainingLabels = samples.loadLabelsFile("digitdata/traininglabels",
numTraining)
print("Extracting features...")
testData = map(featureFunction, rawTestData)
validationData = map(featureFunction, rawValidationData)
trainingData = map(featureFunction, rawTrainingData)
for percent in range(1, 11):
acc = []
aTime = []
for runCount in range(0, 4):
# Extract features
print("======================================")
print("(" + str(runCount) + ")", "Building random",
(percent * 10), " percent of Training Data...")
numSubTraining = int((percent / 10.0) * totalTrainData)
indexes = random.sample(range(0, totalTrainData),
numSubTraining)
subTrainingData = []
subTrainingLabels = []
for indx in indexes:
subTrainingData.append(trainingData[indx])
subTrainingLabels.append(trainingLabels[indx])
validationData = []
validationLabels = []
# Conduct training and testing
start = time.time()
print("(" + str(runCount) + ")", "Training", numSubTraining,
"points ...")
if options.classifier == 'nearestNeighbors':
classifier.train(trainingData, trainingLabels, testData,
testLabels, options.k_number_of_neighbors)
else:
classifier.train(subTrainingData, subTrainingLabels,
validationData, validationLabels)
end = time.time()
elapsed = end - start
print("Elapsed Time:", elapsed)
aTime.append(elapsed)
print("(" + str(runCount) + ")", "Testing...")
guesses = classifier.classify(testData)
correct = [
guesses[i] == testLabels[i] for i in range(len(testLabels))
].count(True)
print(
str(correct),
("correct out of " + str(len(testLabels)) + " (%.1f%%).") %
(100.0 * correct / len(testLabels)))
acc.append(100.0 * correct / len(testLabels))
mean = 0
avgT = 0
for q in range(0, len(acc)):
mean += acc[q]
avgT += aTime[q]
mean = mean / len(acc)
avgT = avgT / len(aTime)
print("---------------")
print("Average training time for", numSubTraining, "data points: ",
avgT)
print("Average accuracy of", (percent * 10),
"percent data training: ", str(mean))
sd = 0
for a in acc:
tmp = a - mean
sd += (tmp * tmp)
sd = sd / (len(acc) - 1)
sd = math.sqrt(sd)
print("Standard Derivation in accuracy:", sd)
sys.exit(1)
else:
numTraining = options.training
numTest = options.test
if options.data == "faces":
rawTrainingData = samples.loadDataFile("facedata/facedatatrain",
numTraining,
FACE_DATUM_WIDTH,
FACE_DATUM_HEIGHT)
trainingLabels = samples.loadLabelsFile(
"facedata/facedatatrainlabels", numTraining)
rawValidationData = samples.loadDataFile("facedata/facedatatrain",
numTest, FACE_DATUM_WIDTH,
FACE_DATUM_HEIGHT)
validationLabels = samples.loadLabelsFile(
"facedata/facedatatrainlabels", numTest)
rawTestData = samples.loadDataFile("facedata/facedatatest",
numTest, FACE_DATUM_WIDTH,
FACE_DATUM_HEIGHT)
testLabels = samples.loadLabelsFile("facedata/facedatatestlabels",
numTest)
else:
rawTrainingData = samples.loadDataFile("digitdata/trainingimages",
numTraining,
DIGIT_DATUM_WIDTH,
DIGIT_DATUM_HEIGHT)
trainingLabels = samples.loadLabelsFile("digitdata/traininglabels",
numTraining)
rawValidationData = samples.loadDataFile(
"digitdata/validationimages", numTest, DIGIT_DATUM_WIDTH,
DIGIT_DATUM_HEIGHT)
validationLabels = samples.loadLabelsFile(
"digitdata/validationlabels", numTest)
rawTestData = samples.loadDataFile("digitdata/testimages", numTest,
DIGIT_DATUM_WIDTH,
DIGIT_DATUM_HEIGHT)
testLabels = samples.loadLabelsFile("digitdata/testlabels",
numTest)
# Extract features
print("Extracting features...")
trainingData = map(featureFunction, rawTrainingData)
validationData = map(featureFunction, rawValidationData)
testData = map(featureFunction, rawTestData)
if options.k_number_of_neighbors > 0:
k = options.k_number_of_neighbors
# Conduct training and testing
print("Training...")
if options.classifier == 'nearestNeighbors':
classifier.train(trainingData, trainingLabels, testData,
testLabels, options.k_number_of_neighbors)
else:
classifier.train(trainingData, trainingLabels, validationData,
validationLabels)
if options.classifier != 'nearestNeighbors':
print("Validating...")
guesses = classifier.classify(validationData)
correct = [
guesses[i] == validationLabels[i]
for i in range(len(validationLabels))
].count(True)
print(str(correct),
("correct out of " + str(len(validationLabels)) +
" (%.1f%%).") % (100.0 * correct / len(validationLabels)))
print("Testing...")
guesses = classifier.classify(testData)
correct = [
guesses[i] == testLabels[i] for i in range(len(testLabels))
].count(True)
print(str(correct),
("correct out of " + str(len(testLabels)) + " (%.1f%%).") %
(100.0 * correct / len(testLabels)))
analysis(classifier, guesses, testLabels, testData, rawTestData,
printImage)
if options.analysis:
analysis(classifier, guesses, testLabels, testData, rawTestData,
printImage)
# do odds ratio computation if specified at command line
if options.odds & (options.classifier == "naiveBayes" or
(options.classifier == "nb")):
label1, label2 = options.label1, options.label2
features_odds = classifier.findHighOddsFeatures(label1, label2)
if options.classifier == "naiveBayes" or options.classifier == "nb":
string3 = "=== Features with highest odd ratio of label %d over label %d ===" % (
label1, label2)
else:
string3 = "=== Features for which weight(label %d)-weight(label %d) is biggest ===" % (
label1, label2)
print(string3)
printImage(features_odds)
if options.weights & (options.classifier == "perceptron"):
for l in classifier.legalLabels:
features_weights = classifier.findHighWeightFeatures(l)
print("=== Features with high weight for label %d ===" % l)
printImage(features_weights)
def runClassifier(args, options):
#print 'args: ', args
#print 'options', options
featureFunction = args['featureFunction']
classifier = args['classifier']
printImage = args['printImage']
# Load data
numTraining = options.training
numTest = options.test
if(options.data=="faces"):
rawTrainingData = samples.loadDataFile("facedata/facedatatrain", numTraining,FACE_DATUM_WIDTH,FACE_DATUM_HEIGHT)
trainingLabels = samples.loadLabelsFile("facedata/facedatatrainlabels", numTraining)
rawValidationData = samples.loadDataFile("facedata/facedatatrain", numTest,FACE_DATUM_WIDTH,FACE_DATUM_HEIGHT)
validationLabels = samples.loadLabelsFile("facedata/facedatatrainlabels", numTest)
rawTestData = samples.loadDataFile("facedata/facedatatest", numTest,FACE_DATUM_WIDTH,FACE_DATUM_HEIGHT)
testLabels = samples.loadLabelsFile("facedata/facedatatestlabels", numTest)
else:
rawTrainingData = samples.loadDataFile("digitdata/trainingimages", numTraining,DIGIT_DATUM_WIDTH,DIGIT_DATUM_HEIGHT)
trainingLabels = samples.loadLabelsFile("digitdata/traininglabels", numTraining)
rawValidationData = samples.loadDataFile("digitdata/validationimages", numTest,DIGIT_DATUM_WIDTH,DIGIT_DATUM_HEIGHT)
validationLabels = samples.loadLabelsFile("digitdata/validationlabels", numTest)
rawTestData = samples.loadDataFile("digitdata/testimages", numTest,DIGIT_DATUM_WIDTH,DIGIT_DATUM_HEIGHT)
testLabels = samples.loadLabelsFile("digitdata/testlabels", numTest)
# Extract features
#print "Extracting features..."
#print '#######type of rawTrainingData is', rawTrainingData.__class__ # list of Datum
#print '#######type of rawTrainingData[0] is', rawTrainingData[0].__class__ # Datum
trainingData = map(featureFunction, rawTrainingData)
#print '#######type of trainingData is', trainingData.__class__ # list of Counter
#print '#######type of trainingData[0] is', trainingData[0].__class__ # Counter
validationData = map(featureFunction, rawValidationData)
testData = map(featureFunction, rawTestData)
# Conduct training and testing
print "Training..."
classifier.train(trainingData, trainingLabels, validationData, validationLabels)
print "Validating..."
guesses = classifier.classify(validationData)
correct = [guesses[i] == validationLabels[i] for i in range(len(validationLabels))].count(True)
print str(correct), ("correct out of " + str(len(validationLabels)) + " (%.1f%%).") % (100.0 * correct / len(validationLabels))
print "Testing..."
guesses = classifier.classify(testData)
correct = [guesses[i] == testLabels[i] for i in range(len(testLabels))].count(True)
print str(correct), ("correct out of " + str(len(testLabels)) + " (%.1f%%).") % (100.0 * correct / len(testLabels))
analysis(classifier, guesses, testLabels, testData, rawTestData, printImage)
# do odds ratio computation if specified at command line
if((options.odds) & (options.classifier == "naiveBayes" or (options.classifier == "nb")) ):
label1, label2 = options.label1, options.label2
features_odds = classifier.findHighOddsFeatures(label1,label2)
if(options.classifier == "naiveBayes" or options.classifier == "nb"):
string3 = "=== Features with highest odd ratio of label %d over label %d ===" % (label1, label2)
else:
string3 = "=== Features for which weight(label %d)-weight(label %d) is biggest ===" % (label1, label2)
print string3
printImage(features_odds)
if((options.weights) & (options.classifier == "perceptron")):
for l in classifier.legalLabels:
features_weights = classifier.findHighWeightFeatures(l)
print ("=== Features with high weight for label %d ==="%l)
printImage(features_weights)
def runClassifier(args, options):
featureFunction = args['featureFunction']
classifier = args['classifier']
printImage = args['printImage']
# Load data
numTraining = options.training
numTest = options.test
thisWidth = 0
thisHeight = 0
num = 0
sample = options.sample
reps = options.repetitions
acc = []
stdDev = []
for a in range(sample):
acc.append(0)
counter = 0
for r in range(reps):
print("______Repetition " + str(r + 1) + " out of " + str(reps) +
"______")
counter += 1
accTemp = []
for i in range(sample):
print("__________" + str(i + 1) + " try with " +
str(100 * (i + 1) / sample) +
"% percent of random training data__________")
arr = [j for j in range(numTraining)]
random.shuffle(arr)
if (options.data == "faces"):
rawTrainingData1 = samples.loadDataFile(
"facedata/facedatatrain", numTraining, FACE_DATUM_WIDTH,
FACE_DATUM_HEIGHT)
trainingLabels1 = samples.loadLabelsFile(
"facedata/facedatatrainlabels", numTraining)
rawValidationData = samples.loadDataFile(
"facedata/facedatatrain", numTest, FACE_DATUM_WIDTH,
FACE_DATUM_HEIGHT)
validationLabels = samples.loadLabelsFile(
"facedata/facedatatrainlabels", numTest)
rawTestData = samples.loadDataFile("facedata/facedatatest",
numTest, FACE_DATUM_WIDTH,
FACE_DATUM_HEIGHT)
testLabels = samples.loadLabelsFile(
"facedata/facedatatestlabels", numTest)
thisWidth = FACE_DATUM_WIDTH
thisHeight = FACE_DATUM_HEIGHT
num = 2
else:
rawTrainingData1 = samples.loadDataFile(
"digitdata/trainingimages", numTraining, DIGIT_DATUM_WIDTH,
DIGIT_DATUM_HEIGHT)
trainingLabels1 = samples.loadLabelsFile(
"digitdata/traininglabels", numTraining)
rawValidationData = samples.loadDataFile(
"digitdata/validationimages", numTest, DIGIT_DATUM_WIDTH,
DIGIT_DATUM_HEIGHT)
validationLabels = samples.loadLabelsFile(
"digitdata/validationlabels", numTest)
rawTestData = samples.loadDataFile("digitdata/testimages",
numTest, DIGIT_DATUM_WIDTH,
DIGIT_DATUM_HEIGHT)
testLabels = samples.loadLabelsFile("digitdata/testlabels",
numTest)
thisWidth = DIGIT_DATUM_WIDTH
thisHeight = DIGIT_DATUM_HEIGHT
num = 10
rawTrainingData = []
trainingLabels = []
for n in arr[:(numTraining * (1 + i) / sample)]:
rawTrainingData.append(rawTrainingData1[n])
trainingLabels.append(trainingLabels1[n])
# Extract features
print "Extracting features..."
trainingData = map(featureFunction, rawTrainingData)
validationData = map(featureFunction, rawValidationData)
testData = map(featureFunction, rawTestData)
# Conduct training and testing
print "Training..."
classifier.train(trainingData, trainingLabels, validationData,
validationLabels, thisWidth, thisHeight, num)
print "Validating..."
guesses = classifier.classify(validationData)
correct = [
guesses[i] == validationLabels[i]
for i in range(len(validationLabels))
].count(True)
print str(correct), ("correct out of " +
str(len(validationLabels)) + " (%.1f%%).") % (
100.0 * correct / len(validationLabels))
print "Testing..."
guesses = classifier.classify(testData)
correct = [
guesses[i] == testLabels[i] for i in range(len(testLabels))
].count(True)
percentCorrect = (100.0 * correct / len(testLabels))
print str(correct), ("correct out of " + str(len(testLabels)) +
" (" + str(percentCorrect) + "%).")
accTemp.append(percentCorrect)
#analysis(classifier, guesses, testLabels, testData, rawTestData, printImage)
for i in range(len(acc)):
acc[i] += accTemp[i]
stdDev.append(accTemp)
accAvg = []
stdDev2 = [[] for k in range(sample)]
for i in range(sample):
for j in range(reps):
stdDev2[i].append(stdDev[j][i])
for i in range(len(acc)):
accAvg.append(acc[i] / reps)
currSTD = np.std(stdDev2[i])
print("Accuracy with " + str(100 * (i + 1) / sample) +
"% of training data: " + str(accAvg[i]) + "%, Standard Dev: " +
str(currSTD))
print("Repetitions: " + str(reps))
# do odds ratio computation if specified at command line
if ((options.odds) & (options.classifier == "naiveBayes" or
(options.classifier == "nb"))):
label1, label2 = options.label1, options.label2
features_odds = classifier.findHighOddsFeatures(label1, label2)
if (options.classifier == "naiveBayes" or options.classifier == "nb"):
string3 = "=== Features with highest odd ratio of label %d over label %d ===" % (
label1, label2)
else:
string3 = "=== Features for which weight(label %d)-weight(label %d) is biggest ===" % (
label1, label2)
print string3
printImage(features_odds)
if ((options.weights) & (options.classifier == "perceptron")):
for l in classifier.legalLabels:
features_weights = classifier.findHighWeightFeatures(l)
print("=== Features with high weight for label %d ===" % l)
printImage(features_weights)
def runClassifier(args, options):
start = timeit.default_timer()
featureFunction = args['featureFunction']
classifier = args['classifier']
printImage = args['printImage']
percent = args['percent'] / 100.0
# Load data
numTraining = options.training
numTest = options.test
if (options.data == "faces"):
face_test_size = 150
rawTrainingData = samples.loadDataFile("facedata/facedatatrain", 450, FACE_DATUM_WIDTH,
FACE_DATUM_HEIGHT)
trainingLabels = samples.loadLabelsFile("facedata/facedatatrainlabels", 450)
zipped = list(zip(rawTrainingData, trainingLabels));
amount = int(450 * percent)
randomed = sample(zipped, amount)
unzipped = zip(*randomed)
rawTrainingData = unzipped[0]
trainingLabels = unzipped[1]
rawValidationData = samples.loadDataFile("facedata/facedatatrain", face_test_size, FACE_DATUM_WIDTH, FACE_DATUM_HEIGHT)
validationLabels = samples.loadLabelsFile("facedata/facedatatrainlabels", face_test_size)
rawTestData = samples.loadDataFile("facedata/facedatatest", face_test_size, FACE_DATUM_WIDTH, FACE_DATUM_HEIGHT)
testLabels = samples.loadLabelsFile("facedata/facedatatestlabels", face_test_size)
else:
digit_test_size = 1000
rawTrainingData = samples.loadDataFile("digitdata/trainingimages", 5000, DIGIT_DATUM_WIDTH,
DIGIT_DATUM_HEIGHT)
trainingLabels = samples.loadLabelsFile("digitdata/traininglabels", 5000)
zipped = list(zip(rawTrainingData,trainingLabels));
amount = int(5000 * percent)
randomed = sample(zipped, amount)
unzipped = zip(*randomed)
rawTrainingData = unzipped[0]
trainingLabels = unzipped[1]
print("Len of sampled data + ", len(rawTrainingData), len(trainingLabels))
rawValidationData = samples.loadDataFile("digitdata/validationimages", digit_test_size, DIGIT_DATUM_WIDTH,
DIGIT_DATUM_HEIGHT)
validationLabels = samples.loadLabelsFile("digitdata/validationlabels", digit_test_size)
rawTestData = samples.loadDataFile("digitdata/testimages", digit_test_size, DIGIT_DATUM_WIDTH, DIGIT_DATUM_HEIGHT)
testLabels = samples.loadLabelsFile("digitdata/testlabels", digit_test_size)
# Extract features
print "Extracting features..."
trainingData = map(featureFunction, rawTrainingData)
validationData = map(featureFunction, rawValidationData)
testData = map(featureFunction, rawTestData)
# Conduct training and testing
print "Training..."
classifier.train(trainingData, trainingLabels, validationData, validationLabels)
"""print "Validating..."
guesses = classifier.classify(validationData)
correct = [guesses[i] == validationLabels[i] for i in range(len(validationLabels))].count(True)
print str(correct), ("correct out of " + str(len(validationLabels)) + " (%.1f%%).") % (
100.0 * correct / len(validationLabels))"""
print "Testing..."
guesses = classifier.classify(testData)
correct = [guesses[i] == testLabels[i] for i in range(len(testLabels))].count(True)
print str(correct), ("correct out of " + str(len(testLabels)) + " (%.1f%%).") % (100.0 * correct / len(testLabels))
analysis(classifier, guesses, testLabels, testData, rawTestData, printImage)
# do odds ratio computation if specified at command line
if ((options.odds) & (options.classifier == "naiveBayes" or (options.classifier == "nb"))):
label1, label2 = options.label1, options.label2
features_odds = classifier.findHighOddsFeatures(label1, label2)
if (options.classifier == "naiveBayes" or options.classifier == "nb"):
string3 = "=== Features with highest odd ratio of label %d over label %d ===" % (label1, label2)
else:
string3 = "=== Features for which weight(label %d)-weight(label %d) is biggest ===" % (label1, label2)
print string3
printImage(features_odds)
if ((options.weights) & (options.classifier == "perceptron")):
for l in classifier.legalLabels:
features_weights = classifier.findHighWeightFeatures(l)
print ("=== Features with high weight for label %d ===" % l)
printImage(features_weights)
stop = timeit.default_timer()
print('Classifier Run Time (in seconds): ', stop - start)
import samples
import util
import numpy as np
import os
from samples import Datum
from samples import readlines
from dataClassifier import DIGIT_DATUM_HEIGHT, DIGIT_DATUM_WIDTH, contestFeatureExtractorDigit
from samples import IntegerConversionFunction
featureFunction = contestFeatureExtractorDigit
rawTrainingData = samples.loadDataFile("digitdata/testimages", 1, 28, 28)
trainingLabels = samples.loadLabelsFile("digitdata/traininglabels", 1)
fin = readlines("digitdata/testimages")
fin.reverse()
a = ['+', ' ', '#']
print(IntegerConversionFunction(a))
data = []
items = []
for j in range(28):
data.append(list(fin.pop()))
for i in range(28):
print(data[i])
items.append(Datum(data, 28, 28))
print(items[0].getPixels())
def runClassifier(args, options):
featureFunction = args['featureFunction']
classifier = args['classifier']
printImage = args['printImage']
# Load data
numTraining = options.training
numTest = options.test
percent = options.percentage
if (options.data == "faces"):
rawTrainingData = samples.loadDataFile("facedata/facedatatrain",
numTraining, FACE_DATUM_WIDTH,
FACE_DATUM_HEIGHT)
trainingLabels = samples.loadLabelsFile("facedata/facedatatrainlabels",
numTraining)
rawValidationData = samples.loadDataFile("facedata/facedatatrain",
numTest, FACE_DATUM_WIDTH,
FACE_DATUM_HEIGHT)
validationLabels = samples.loadLabelsFile(
"facedata/facedatatrainlabels", numTest)
rawTestData = samples.loadDataFile("facedata/facedatatest", numTest,
FACE_DATUM_WIDTH, FACE_DATUM_HEIGHT)
testLabels = samples.loadLabelsFile("facedata/facedatatestlabels",
numTest)
else:
rawTrainingData = samples.loadDataFile("digitdata/trainingimages",
numTraining, DIGIT_DATUM_WIDTH,
DIGIT_DATUM_HEIGHT)
trainingLabels = samples.loadLabelsFile("digitdata/traininglabels",
numTraining)
rawValidationData = samples.loadDataFile("digitdata/validationimages",
numTest, DIGIT_DATUM_WIDTH,
DIGIT_DATUM_HEIGHT)
validationLabels = samples.loadLabelsFile("digitdata/validationlabels",
numTest)
rawTestData = samples.loadDataFile("digitdata/testimages", numTest,
DIGIT_DATUM_WIDTH,
DIGIT_DATUM_HEIGHT)
testLabels = samples.loadLabelsFile("digitdata/testlabels", numTest)
# Extract features
print "Extracting features..."
trainingData = map(featureFunction, rawTrainingData)
validationData = map(featureFunction, rawValidationData)
testData = map(featureFunction, rawTestData)
if options.k_number_of_neighbors > 0:
k = options.k_number_of_neighbors
# Conduct training and testing
print "Training..."
if (options.classifier == 'nearestNeighbors'):
classifier.train(trainingData, trainingLabels, testData, testLabels,
options.k_number_of_neighbors)
else:
randTrainingData, randTrainingLabels = randomData(
trainingData, trainingLabels, percent)
start = time.clock()
classifier.train(randTrainingData, randTrainingLabels, validationData,
validationLabels)
runTime = time.clock() - start
print "training set runtime:\t" + str(runTime)
if (options.classifier != 'nearestNeighbors'):
print "Validating..."
guesses = classifier.classify(validationData)
correct = [
guesses[i] == validationLabels[i]
for i in range(len(validationLabels))
].count(True)
print str(correct), ("correct out of " + str(len(validationLabels)) +
" (%.1f%%).") % (100.0 * correct /
len(validationLabels))
print "Testing..."
guesses = classifier.classify(testData)
correct = [
guesses[i] == testLabels[i] for i in range(len(testLabels))
].count(True)
print str(correct), ("correct out of " + str(len(testLabels)) +
" (%.1f%%).") % (100.0 * correct /
len(testLabels))
# analysis(classifier, guesses, testLabels, testData, rawTestData, printImage)
# do odds ratio computation if specified at command line
if ((options.odds) & (options.classifier == "naiveBayes" or
(options.classifier == "nb"))):
label1, label2 = options.label1, options.label2
features_odds = classifier.findHighOddsFeatures(label1, label2)
if (options.classifier == "naiveBayes" or options.classifier == "nb"):
string3 = "=== Features with highest odd ratio of label %d over label %d ===" % (
label1, label2)
else:
string3 = "=== Features for which weight(label %d)-weight(label %d) is biggest ===" % (
label1, label2)
print string3
printImage(features_odds)
if ((options.weights) & (options.classifier == "perceptron")):
for l in classifier.legalLabels:
features_weights = classifier.findHighWeightFeatures(l)
print("=== Features with high weight for label %d ===" % l)
printImage(features_weights)
import math
import samples
#Naive Bayes - Face Data
if __name__ == '__main__':
print "Training Phase"
#stores training data and appropriate labels for faces
n = 450
items = samples.loadDataFile("facedata/facedatatrain", n, 60, 70)
labels = samples.loadLabelsFile("facedata/facedatatrainlabels", n)
all_feature_vectors = [] #stores all quadrants of all sample images
for k in range(n):
#break up face data into 100 6x7 pixel quadrants for feature extraction
feature_quadrants = [] #will be a list of lists
temp_array = []
i_start = 0
i_end = 6
j_start = 0
j_end = 7
while i_end <= 60 and j_end <= 70:
#parse through image and store pixels in a temporary array
for i in range(i_start, i_end):
for j in range(j_start, j_end):
temp_array.append(items[k].getPixel(i, j))
#add temp_array to feature_quadrant array and reassign temp_array
feature_quadrants.append(temp_array)
temp_array = []
def runClassifier(args, options):
featureFunction = args['featureFunction']
classifier = args['classifier']
printImage = args['printImage']
# Load data
numTraining = options.training
numTest = options.test
if (options.data == "pacman"):
agentToClone = args.get('agentToClone', None)
trainingData, validationData, testData = MAP_AGENT_TO_PATH_OF_SAVED_GAMES.get(
agentToClone, (None, None, None))
trainingData = trainingData or args.get(
'trainingData',
False) or MAP_AGENT_TO_PATH_OF_SAVED_GAMES['ContestAgent'][0]
validationData = validationData or args.get(
'validationData',
False) or MAP_AGENT_TO_PATH_OF_SAVED_GAMES['ContestAgent'][1]
testData = testData or MAP_AGENT_TO_PATH_OF_SAVED_GAMES[
'ContestAgent'][2]
rawTrainingData, trainingLabels = samples.loadPacmanData(
trainingData, numTraining)
rawValidationData, validationLabels = samples.loadPacmanData(
validationData, numTest)
rawTestData, testLabels = samples.loadPacmanData(testData, numTest)
else:
rawTrainingData = samples.loadDataFile("digitdata/trainingimages",
numTraining, DIGIT_DATUM_WIDTH,
DIGIT_DATUM_HEIGHT)
trainingLabels = samples.loadLabelsFile("digitdata/traininglabels",
numTraining)
rawValidationData = samples.loadDataFile("digitdata/validationimages",
numTest, DIGIT_DATUM_WIDTH,
DIGIT_DATUM_HEIGHT)
validationLabels = samples.loadLabelsFile("digitdata/validationlabels",
numTest)
rawTestData = samples.loadDataFile("digitdata/testimages", numTest,
DIGIT_DATUM_WIDTH,
DIGIT_DATUM_HEIGHT)
testLabels = samples.loadLabelsFile("digitdata/testlabels", numTest)
# Extract features
print "Extracting features..."
trainingData = map(featureFunction, rawTrainingData)
validationData = map(featureFunction, rawValidationData)
testData = map(featureFunction, rawTestData)
# Conduct training and testing
print "Training..."
classifier.train(trainingData, trainingLabels, validationData,
validationLabels)
print "Validating..."
guesses = classifier.classify(validationData)
correct = [
guesses[i] == validationLabels[i] for i in range(len(validationLabels))
].count(True)
print str(correct), ("correct out of " + str(len(validationLabels)) +
" (%.1f%%).") % (100.0 * correct /
len(validationLabels))
print "Testing..."
guesses = classifier.classify(testData)
correct = [guesses[i] == testLabels[i]
for i in range(len(testLabels))].count(True)
print str(correct), ("correct out of " + str(len(testLabels)) +
" (%.1f%%).") % (100.0 * correct / len(testLabels))
def runClassifier(args, options):
featureFunction = args['featureFunction']
classifier = args['classifier']
printImage = args['printImage']
# Load data
numTraining = options.training
numTest = options.test
rawTrainingData = samples.loadDataFile("digitdata/trainingimages",
numTraining, DIGIT_DATUM_WIDTH,
DIGIT_DATUM_HEIGHT)
trainingLabels = samples.loadLabelsFile("digitdata/traininglabels",
numTraining)
rawValidationData = samples.loadDataFile("digitdata/validationimages",
numTest, DIGIT_DATUM_WIDTH,
DIGIT_DATUM_HEIGHT)
validationLabels = samples.loadLabelsFile("digitdata/validationlabels",
numTest)
rawTestData = samples.loadDataFile("digitdata/testimages", numTest,
DIGIT_DATUM_WIDTH, DIGIT_DATUM_HEIGHT)
testLabels = samples.loadLabelsFile("digitdata/testlabels", numTest)
# Extract features
print "Extracting features..."
trainingData = map(featureFunction, rawTrainingData)
validationData = map(featureFunction, rawValidationData)
testData = map(featureFunction, rawTestData)
# Conduct training and testing
print "Training..."
classifier.train(trainingData, trainingLabels, validationData,
validationLabels, options.validate)
print "Validating..."
guesses = classifier.classify(validationData)
correct = [
guesses[i] == validationLabels[i] for i in range(len(validationLabels))
].count(True)
print str(correct), ("correct out of " + str(len(validationLabels)) +
" (%.1f%%).") % (100.0 * correct /
len(validationLabels))
if (options.classifier == "perceptron"):
f = open("perceptron_valid.csv", "a")
f.write(
str(len(trainingData)) + "," +
str(100 * correct / (1.0 * (len(validationData)))) + '\n')
f.close()
print "Testing..."
guesses = classifier.classify(testData)
correct = [guesses[i] == testLabels[i]
for i in range(len(testLabels))].count(True)
print str(correct), ("correct out of " + str(len(testLabels)) +
" (%.1f%%).") % (100.0 * correct / len(testLabels))
analysis(classifier, guesses, testLabels, testData, rawTestData,
printImage)
if (options.classifier == "perceptron"):
f = open("perceptron_test.csv", "a")
f.write(
str(len(trainingData)) + "," + str(100 * correct /
(1.0 * (len(testData)))) + '\n')
f.close()
def runClassifier(args, options):
featureFunction = args['featureFunction']
classifier = args['classifier']
printImage = args['printImage']
# Load data
numTraining = options.training
# Extract features
print "Extracting features..."
if options.data == "faces":
rawTrainingData = samples.loadDataFile("facedata/facedatatrain",
numTraining, FACE_DATUM_WIDTH,
FACE_DATUM_HEIGHT)
trainingLabels = samples.loadLabelsFile("facedata/facedatatrainlabels",
numTraining)
rawValidationData = samples.loadDataFile("facedata/facedatavalidation",
TEST_SET_SIZE,
FACE_DATUM_WIDTH,
FACE_DATUM_HEIGHT)
validationLabels = samples.loadLabelsFile(
"facedata/facedatavalidationlabels", TEST_SET_SIZE)
rawTestData = samples.loadDataFile("facedata/facedatatest",
TEST_SET_SIZE, FACE_DATUM_WIDTH,
FACE_DATUM_HEIGHT)
testLabels = samples.loadLabelsFile("facedata/facedatatestlabels",
TEST_SET_SIZE)
else:
rawTrainingData = samples.loadDataFile("digitdata/trainingimages",
numTraining, DIGIT_DATUM_WIDTH,
DIGIT_DATUM_HEIGHT)
trainingLabels = samples.loadLabelsFile("digitdata/traininglabels",
numTraining)
rawValidationData = samples.loadDataFile("digitdata/validationimages",
TEST_SET_SIZE,
DIGIT_DATUM_WIDTH,
DIGIT_DATUM_HEIGHT)
validationLabels = samples.loadLabelsFile("digitdata/validationlabels",
TEST_SET_SIZE)
rawTestData = samples.loadDataFile("digitdata/testimages",
TEST_SET_SIZE, DIGIT_DATUM_WIDTH,
DIGIT_DATUM_HEIGHT)
testLabels = samples.loadLabelsFile("digitdata/testlabels",
TEST_SET_SIZE)
if options.classifier == "GDA":
if options.data == "faces":
dimension = 13
principleComponents = getPrincipleComponents(
map(
featureFunction,
samples.loadDataFile("facedata/facedatatrain", 451,
FACE_DATUM_WIDTH, FACE_DATUM_HEIGHT)),
dimension)
trainingData = np.dot(
basicFeatureDataToNumpyArray(
map(featureFunction, rawTrainingData)),
principleComponents)
validationData = np.dot(
basicFeatureDataToNumpyArray(
map(featureFunction, rawValidationData)),
principleComponents)
testData = np.dot(
basicFeatureDataToNumpyArray(map(featureFunction,
rawTestData)),
principleComponents)
else:
dimension = 13
principleComponents = getPrincipleComponents(
map(
featureFunction,
samples.loadDataFile("digitdata/trainingimages", 5000,
DIGIT_DATUM_WIDTH,
DIGIT_DATUM_HEIGHT)), dimension)
trainingData = np.dot(
basicFeatureDataToNumpyArray(
map(featureFunction, rawTrainingData)),
principleComponents)
validationData = np.dot(
basicFeatureDataToNumpyArray(
map(featureFunction, rawValidationData)),
principleComponents)
testData = np.dot(
basicFeatureDataToNumpyArray(map(featureFunction,
rawTestData)),
principleComponents)
else:
trainingData = map(featureFunction, rawTrainingData)
validationData = map(featureFunction, rawValidationData)
testData = map(featureFunction, rawTestData)
# Conduct training and testing
print "Training..."
classifier.train(trainingData, trainingLabels, validationData,
validationLabels)
print "Validating..."
guesses = classifier.classify(validationData)
correct = [
guesses[i] == validationLabels[i] for i in range(len(validationLabels))
].count(True)
print str(correct), ("correct out of " + str(len(validationLabels)) +
" (%.1f%%).") % (100.0 * correct /
len(validationLabels))
print "Testing..."
guesses = classifier.classify(testData)
correct = [guesses[i] == testLabels[i]
for i in range(len(testLabels))].count(True)
print str(correct), ("correct out of " + str(len(testLabels)) +
" (%.1f%%).") % (100.0 * correct / len(testLabels))
analysis(classifier, guesses, testLabels, testData, rawTestData,
printImage)
def runClassifier(args, options):
featureFunction = args['featureFunction']
classifier = args['classifier']
printImage = args['printImage']
# Load data
numTraining = options.training
if (options.data == "faces"):
rawTrainingData = samples.loadDataFile("facedata/facedatatrain",
numTraining, FACE_DATUM_WIDTH,
FACE_DATUM_HEIGHT)
trainingLabels = samples.loadLabelsFile("facedata/facedatatrainlabels",
numTraining)
rawValidationData = samples.loadDataFile("facedata/facedatatrain",
TEST_SET_SIZE,
FACE_DATUM_WIDTH,
FACE_DATUM_HEIGHT)
validationLabels = samples.loadLabelsFile(
"facedata/facedatatrainlabels", TEST_SET_SIZE)
rawTestData = samples.loadDataFile("facedata/facedatatest",
TEST_SET_SIZE, FACE_DATUM_WIDTH,
FACE_DATUM_HEIGHT)
testLabels = samples.loadLabelsFile("facedata/facedatatestlabels",
TEST_SET_SIZE)
else:
rawTrainingData = samples.loadDataFile("digitdata/trainingimages",
numTraining, DIGIT_DATUM_WIDTH,
DIGIT_DATUM_HEIGHT)
trainingLabels = samples.loadLabelsFile("digitdata/traininglabels",
numTraining)
rawValidationData = samples.loadDataFile("digitdata/validationimages",
TEST_SET_SIZE,
DIGIT_DATUM_WIDTH,
DIGIT_DATUM_HEIGHT)
validationLabels = samples.loadLabelsFile("digitdata/validationlabels",
TEST_SET_SIZE)
rawTestData = samples.loadDataFile("digitdata/testimages",
TEST_SET_SIZE, DIGIT_DATUM_WIDTH,
DIGIT_DATUM_HEIGHT)
testLabels = samples.loadLabelsFile("digitdata/testlabels",
TEST_SET_SIZE)
# Extract features
print "Extracting features..."
trainingData = map(featureFunction, rawTrainingData)
validationData = map(featureFunction, rawValidationData)
testData = map(featureFunction, rawTestData)
# Conduct training and testing
print "Training..."
classifier.train(trainingData, trainingLabels, validationData,
validationLabels)
print "Validating..."
guesses = classifier.classify(validationData)
print "len guess: %d valid: %d" % (len(guesses), len(validationLabels))
print guesses[0], guesses[10]
correct = [
guesses[i] == validationLabels[i] for i in range(len(validationLabels))
].count(True)
print str(correct), ("correct out of " + str(len(validationLabels)) +
" (%.1f%%).") % (100.0 * correct /
len(validationLabels))
print "Testing..."
guesses = classifier.classify(testData)
correct = [guesses[i] == testLabels[i]
for i in range(len(testLabels))].count(True)
print str(correct), ("correct out of " + str(len(testLabels)) +
" (%.1f%%).") % (100.0 * correct / len(testLabels))
analysis(classifier, guesses, testLabels, testData, rawTestData,
printImage)
# do odds ratio computation if specified at command line
if ((options.odds) & (options.classifier != "mostFrequent")):
label1, label2 = options.label1, options.label2
features_odds = classifier.findHighOddsFeatures(label1, label2)
if (options.classifier == "naiveBayes" or options.classifier == "nb"):
string3 = "=== Features with highest odd ratio of label %d over label %d ===" % (
label1, label2)
else:
string3 = "=== Features for which weight(label %d)-weight(label %d) is biggest ===" % (
label1, label2)
print string3
printImage(features_odds)
def runClassifier(args, options):
featureFunction = args['featureFunction']
classifier = args['classifier']
printImage = args['printImage']
# Load data
numTraining = options.training
numTest = options.test
if(options.data=="faces"):
rawTrainingData = samples.loadDataFile("facedata/facedatatrain", numTraining,FACE_DATUM_WIDTH,FACE_DATUM_HEIGHT)
trainingLabels = samples.loadLabelsFile("facedata/facedatatrainlabels", numTraining)
rawValidationData = samples.loadDataFile("facedata/facedatatrain", numTest,FACE_DATUM_WIDTH,FACE_DATUM_HEIGHT)
validationLabels = samples.loadLabelsFile("facedata/facedatatrainlabels", numTest)
rawTestData = samples.loadDataFile("facedata/facedatatest", numTest,FACE_DATUM_WIDTH,FACE_DATUM_HEIGHT)
testLabels = samples.loadLabelsFile("facedata/facedatatestlabels", numTest)
else:
rawTrainingData = samples.loadDataFile("digitdata/trainingimages", numTraining,DIGIT_DATUM_WIDTH,DIGIT_DATUM_HEIGHT)
trainingLabels = samples.loadLabelsFile("digitdata/traininglabels", numTraining)
rawValidationData = samples.loadDataFile("digitdata/validationimages", numTest,DIGIT_DATUM_WIDTH,DIGIT_DATUM_HEIGHT)
validationLabels = samples.loadLabelsFile("digitdata/validationlabels", numTest)
rawTestData = samples.loadDataFile("digitdata/testimages", numTest,DIGIT_DATUM_WIDTH,DIGIT_DATUM_HEIGHT)
testLabels = samples.loadLabelsFile("digitdata/testlabels", numTest)
# Extract features
print "Extracting features..."
trainingData = map(featureFunction, rawTrainingData)
validationData = map(featureFunction, rawValidationData)
testData = map(featureFunction, rawTestData)
# Conduct training and testing
print "Start training..."
start = time.time()
classifier.train(trainingData, trainingLabels, validationData, validationLabels)
end = time.time() - start
print "Traning time: " + str(end)
print "Start validating..."
guesses = classifier.classify(validationData)
correct = [guesses[i] == validationLabels[i] for i in range(len(validationLabels))].count(True)
print "Validation result: ", str(correct), ("correct out of " + str(len(validationLabels)) + " (%.1f%%).") % (100.0 * correct / len(validationLabels))
print "Start testing..."
guesses = classifier.classify(testData)
correct = [guesses[i] == testLabels[i] for i in range(len(testLabels))].count(True)
print "Testing result: ", str(correct), ("correct out of " + str(len(testLabels)) + " (%.1f%%).") % (100.0 * correct / len(testLabels))
#analysis(classifier, guesses, testLabels, testData, rawTestData, printImage)
# do odds ratio computation if specified at command line
if((options.odds) & (options.classifier == NB) ):
label1, label2 = options.label1, options.label2
features_odds = classifier.findHighOddsFeatures(label1,label2)
if(options.classifier == NB):
string3 = "=== Features with highest odd ratio of label %d over label %d ===" % (label1, label2)
else:
string3 = "=== Features for which weight(label %d)-weight(label %d) is biggest ===" % (label1, label2)
print string3
printImage(features_odds)
if((options.weights) & (options.classifier == PT)):
for l in classifier.legalLabels:
features_weights = classifier.findHighWeightFeatures(l)
print ("=== Features with high weight for label %d ==="%l)
printImage(features_weights)
"""This file is in Beta and is not the real autograder."""
import data_classification_utils as dcu
import samples
import numpy as np
training_data = samples.loadDataFile("digitdata/trainingimages", 1, 28, 28)
features = dcu.simple_image_featurization(training_data[0])
expected = np.array([0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 1., 1., 1., 1., 1., 2., 2., 1., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 1., 1., 1., 1., 1., 2., 2., 2., 2., 2., 2., 1., 2., 2., 2., 1., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 1., 2., 2., 2., 2., 2., 2., 2., 2., 2., 2., 1., 1., 1., 1., 1., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 2., 2., 2., 2., 2., 2., 2., 1., 2., 2., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 1., 1., 1., 2., 2., 2., 0., 0., 1., 1., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 1., 2., 1., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 1., 2., 1., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 1., 2., 1., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 1., 2., 2., 1., 1., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 1., 2., 2., 2., 1., 1., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 1., 1., 2., 2., 1., 1., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 1., 2., 2., 1., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 2., 2., 2., 1., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 1., 1., 1., 2., 2., 2., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 1., 1., 2., 2., 2., 2., 2., 1., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 1., 1., 2., 2., 2., 2., 2., 2., 1., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 1., 1., 2., 2., 2., 2., 2., 2., 1., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 1., 2., 2., 2., 2., 2., 2., 1., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 1., 1., 2., 2., 2., 2., 2., 2., 1., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 1., 2., 2., 2., 2., 1., 1., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.])
if not np.array_equal(features, expected):
print("Error, featurization is incorrect. You reported: ")
print(features)
def runClassifier():
"""
Harness code for running different classifiers on the face or digit data.
This is the main function for classification, and is designed
to be invoked from the command line (outside the Python interpreter).
Usage:
> python dataClassifier.py
OR
> python dataClassifier.py <data> <classifierName>
OR
> python dataClassifier.py <data> <classifierName> <featureFunction>
OR
> python dataClassifier.py <data> <classifierName> <featureFunction> <numTrainingExamples>
OR
> python dataClassifier.py <data> <classifierName> <featureFunction> <numTrainingExamples> <odds class1 class2>
For example:
> python dataClassifier.py digits naivebayes basic 1000
would run the naive Bayes classifier on 1000 training examples using the
basicFeatureExtractor function, and then test the classifier on the test data.
"""
print "Doing classification"
print "--------------------"
# Assign default values for arguments if they are not provided.
if (len(sys.argv) == 1):
print "No data specified; using digits."
sys.argv.append("digits")
if (len(sys.argv) == 2):
print "No classifier specified; using default."
sys.argv.append("mostfrequent")
if (len(sys.argv) == 3):
print "No feature extraction function specified; using default."
sys.argv.append("basic")
if (len(sys.argv) == 4):
print "No training set size specified; using default."
sys.argv.append("100")
if (len(sys.argv) == 5):
print "Not doing odds ratio computation."
sys.argv.append("noodds")
# Set up variables according to the command line input.
print "data:\t\t" + sys.argv[1]
print "classifier:\t\t" + sys.argv[2]
print "feature extractor:\t" + sys.argv[3]
print "training set size:\t" + sys.argv[4]
if ((sys.argv[1] == "digits") & (sys.argv[3] == "basic")):
featureFunction = basicFeatureExtractorDigit
elif ((sys.argv[1] == "faces") & (sys.argv[3] == "basic")):
featureFunction = basicFeatureExtractorFace
elif ((sys.argv[1] == "digits") & (sys.argv[3] == "enhanced")):
featureFunction = enhancedFeatureExtractorDigit
elif ((sys.argv[1] == "faces") & (sys.argv[3] == "enhanced")):
featureFunction = enhancedFeatureExtractorFace
else:
print "Unknown feature function:", sys.argv[2]
return
if (sys.argv[1] == "digits"): # if digits detect
legalLabels = range(10)
else: # if face detect
legalLabels = range(2)
if (sys.argv[2] == "mostfrequent"):
classifier = mostFrequent.MostFrequentClassifier(legalLabels)
elif (sys.argv[2] == "naivebayes"):
classifier = naiveBayes.NaiveBayesClassifier(legalLabels)
elif (sys.argv[2] == "perceptron"):
classifier = perceptron.PerceptronClassifier(legalLabels)
else:
print "Unknown classifier:", sys.argv[2]
return
# Load data
numTraining = int(sys.argv[4])
if (sys.argv[1] == "faces"):
rawTrainingData = samples.loadDataFile("facedata/facedatatrain",
numTraining, FACE_DATUM_WIDTH,
FACE_DATUM_HEIGHT)
trainingLabels = samples.loadLabelsFile("facedata/facedatatrainlabels",
numTraining)
rawValidationData = samples.loadDataFile("facedata/facedatatrain",
TEST_SET_SIZE,
FACE_DATUM_WIDTH,
FACE_DATUM_HEIGHT)
validationLabels = samples.loadLabelsFile(
"facedata/facedatatrainlabels", TEST_SET_SIZE)
rawTestData = samples.loadDataFile("facedata/facedatatest",
TEST_SET_SIZE, FACE_DATUM_WIDTH,
FACE_DATUM_HEIGHT)
testLabels = samples.loadLabelsFile("facedata/facedatatestlabels",
TEST_SET_SIZE)
else:
rawTrainingData = samples.loadDataFile("digitdata/trainingimages",
numTraining, DIGIT_DATUM_WIDTH,
DIGIT_DATUM_HEIGHT)
trainingLabels = samples.loadLabelsFile("digitdata/traininglabels",
numTraining)
rawValidationData = samples.loadDataFile("digitdata/validationimages",
TEST_SET_SIZE,
DIGIT_DATUM_WIDTH,
DIGIT_DATUM_HEIGHT)
validationLabels = samples.loadLabelsFile("digitdata/validationlabels",
TEST_SET_SIZE)
rawTestData = samples.loadDataFile("digitdata/testimages",
TEST_SET_SIZE, DIGIT_DATUM_WIDTH,
DIGIT_DATUM_HEIGHT)
testLabels = samples.loadLabelsFile("digitdata/testlabels",
TEST_SET_SIZE)
# Extract features
print "Extracting features..."
trainingData = map(featureFunction, rawTrainingData)
validationData = map(featureFunction, rawValidationData)
testData = map(featureFunction, rawTestData)
# Conduct training and testing
print "Training..."
classifier.train(trainingData, trainingLabels, validationData,
validationLabels)
print "Validating..."
guesses = classifier.classify(validationData)
correct = [
guesses[i] == validationLabels[i] for i in range(len(validationLabels))
].count(True)
print str(correct), ("correct out of " + str(len(validationLabels)) +
" (%.1f%%).") % (100.0 * correct /
len(validationLabels))
print "Testing..."
guesses = classifier.classify(testData)
correct = [guesses[i] == testLabels[i]
for i in range(len(testLabels))].count(True)
print str(correct), ("correct out of " + str(len(testLabels)) +
" (%.1f%%).") % (100.0 * correct / len(testLabels))
util.pause()
analysis(classifier, guesses, testLabels, rawTestData)
# do odds ratio computation if specified at command line
if ((sys.argv[5] == "odds") & (len(sys.argv) == 8)):
features_class1, features_class2, features_odds = classifier.findHighOddsFeatures(
int(sys.argv[6]), int(sys.argv[7]))
if (sys.argv[1] == "faces"):
printImage(features_class1, FACE_DATUM_WIDTH, FACE_DATUM_HEIGHT)
printImage(features_class2, FACE_DATUM_WIDTH, FACE_DATUM_HEIGHT)
printImage(features_odds, FACE_DATUM_WIDTH, FACE_DATUM_HEIGHT)
else:
printImage(features_class1, DIGIT_DATUM_WIDTH, DIGIT_DATUM_HEIGHT)
printImage(features_class2, DIGIT_DATUM_WIDTH, DIGIT_DATUM_HEIGHT)
printImage(features_odds, DIGIT_DATUM_WIDTH, DIGIT_DATUM_HEIGHT)
def runClassifier(args, options):
featureFunction = args['featureFunction']
classifier = args['classifier']
printImage = args['printImage']
# Load data
numTraining = options['train']
if(options['data']=="faces"):
rawTrainingData = samples.loadDataFile("facedata/facedatatrain", numTraining,FACE_DATUM_WIDTH,FACE_DATUM_HEIGHT)
trainingLabels = samples.loadLabelsFile("facedata/facedatatrainlabels", numTraining)
rawValidationData = samples.loadDataFile("facedata/facedatatrain", TEST_SET_SIZE,FACE_DATUM_WIDTH,FACE_DATUM_HEIGHT)
validationLabels = samples.loadLabelsFile("facedata/facedatatrainlabels", TEST_SET_SIZE)
rawTestData = samples.loadDataFile("facedata/facedatatest", TEST_SET_SIZE,FACE_DATUM_WIDTH,FACE_DATUM_HEIGHT)
testLabels = samples.loadLabelsFile("facedata/facedatatestlabels", TEST_SET_SIZE)
else:
rawTrainingData = samples.loadDataFile("digitdata/trainingimages", numTraining,DIGIT_DATUM_WIDTH,DIGIT_DATUM_HEIGHT)
trainingLabels = samples.loadLabelsFile("digitdata/traininglabels", numTraining)
rawValidationData = samples.loadDataFile("digitdata/validationimages", TEST_SET_SIZE,DIGIT_DATUM_WIDTH,DIGIT_DATUM_HEIGHT)
validationLabels = samples.loadLabelsFile("digitdata/validationlabels", TEST_SET_SIZE)
rawTestData = samples.loadDataFile("digitdata/testimages", TEST_SET_SIZE,DIGIT_DATUM_WIDTH,DIGIT_DATUM_HEIGHT)
testLabels = samples.loadLabelsFile("digitdata/testlabels", TEST_SET_SIZE)
# Extract features
print "Extracting features..."
trainingData = map(featureFunction, rawTrainingData)
validationData = map(featureFunction, rawValidationData)
testData = map(featureFunction, rawTestData)
# Conduct training and testing
print "Training..."
classifier.train(trainingData, trainingLabels, validationData, validationLabels)
print "Validating..."
guesses = classifier.classify(validationData)
correct = [guesses[i] == validationLabels[i] for i in range(len(validationLabels))].count(True)
print str(correct), ("correct out of " + str(len(validationLabels)) + " (%.1f%%).") % (100.0 * correct / len(validationLabels))
print "Testing..."
guesses = classifier.classify(testData)
correct = [guesses[i] == testLabels[i] for i in range(len(testLabels))].count(True)
print str(correct), ("correct out of " + str(len(testLabels)) + " (%.1f%%).") % (100.0 * correct / len(testLabels))
util.pause()
analysis(classifier, guesses, testLabels, testData, rawTestData, printImage)
# do odds ratio computation if specified at command line
if((options['odds']) & (options['classifier'] != "mostfrequent")):
class1, class2 = options['class1'], options['class2']
features_class1,features_class2,features_odds = classifier.findHighOddsFeatures(class1,class2)
if(options['classifier'] == "naivebayes"):
string1 = "=== Features with max P(F_i = 1 | class = %d) ===" % class1
string2 = "=== Features with max P(F_i = 1 | class = %d) ===" % class2
string3 = "=== Features with highest odd ratio of class %d over class %d ===" % (class1, class2)
else:
string1 = "=== Features with largest weight for class %d ===" % class1
string2 = "=== Features with largest weight for class %d ===" % class2
string3 = "=== Features with for which weight(class %d)-weight(class %d) is biggest ===" % (class1, class2)
print string1
printImage(features_class1)
print string2
printImage(features_class2)
print string3
printImage(features_odds)
def get_digit_acc2():
images = samples.loadDataFile('digitdata/trainingimages', 5000, 28, 28)
labels = samples.loadLabelsFile('digitdata/traininglabels', 5000)
accuracies = []
times = []
def convertlabelslist(digit, labelslist):
booleanlist = [False for i in range(len(labelslist))]
for i in range(len(labelslist)):
if labelslist[i] == digit:
booleanlist[i] = True
else:
booleanlist[i] = False
return booleanlist
# For each percentage
for i in range(10):
sample_percentage = float((i + 1) / 10.0)
sample_size = int(math.floor(sample_percentage * float(len(labels))))
joinedlists = list(zip(images, labels))
images_sample = []
labels_sample = []
# Shuffle until sample has all 9 digits
while True:
joinedlists = list(zip(images, labels))
random.shuffle(joinedlists)
images_sample, labels_sample = zip(*joinedlists)
images_sample = images_sample[:sample_size]
labels_sample = labels_sample[:sample_size]
#print(labels_sample)
if 0 in labels_sample and 1 in labels_sample and 2 in labels_sample\
and 3 in labels_sample and 4 in labels_sample and 5 in labels_sample\
and 6 in labels_sample and 7 in labels_sample and 8 in labels_sample\
and 9 in labels_sample:
break
# Have 10 different labels list for each digit
# Convert labelslist into true/false instead of numbers
labelslist0 = convertlabelslist(0, labels_sample)
labelslist1 = convertlabelslist(1, labels_sample)
labelslist2 = convertlabelslist(2, labels_sample)
labelslist3 = convertlabelslist(3, labels_sample)
labelslist4 = convertlabelslist(4, labels_sample)
labelslist5 = convertlabelslist(5, labels_sample)
labelslist6 = convertlabelslist(6, labels_sample)
labelslist7 = convertlabelslist(7, labels_sample)
labelslist8 = convertlabelslist(8, labels_sample)
labelslist9 = convertlabelslist(9, labels_sample)
all_labels = [
labelslist0, labelslist1, labelslist2, labelslist3, labelslist4,
labelslist5, labelslist6, labelslist7, labelslist8, labelslist9
]
start = time.time()
# Compute weight vectors for all digits 0-9
all_weight_vectors = []
for j in range(10):
featureslist = perceptron.compute_features2(images_sample)
weights = perceptron.initialize_weights(28 * 28, 0)
computed_weights = perceptron.compute_weights(
weights, featureslist, all_labels[j])
all_weight_vectors.append((computed_weights))
elapsed = round(time.time() - start, 2)
acc = demo_digits(all_weight_vectors)
print(str((i + 1) * 10) + ' ' + str(elapsed) + ' ' + str(acc))
def runTask(task):
print("Grading task " + str(task))
if task == 2 or task == 5:
print(
"This is a manually graded task, write your answers in the pdf file"
)
elif task == 1:
print("The solution cannot be exposed to you now. :)")
elif task == 3:
print("Ungraded task")
else:
if task == 6:
numTraining = 800
numTest = 200
num_classes = 4
else:
numTraining = dataClassifier.TRAIN_SET_SIZE
numTest = dataClassifier.TEST_SET_SIZE
num_classes = 10
if task == 6:
rawTrainingData = samples.loadDataFile("data/D2/training_data",
numTraining)
trainingLabels = samples.loadLabelsFile("data/D2/training_labels",
numTraining)
rawTestData = samples.loadDataFile("data/D2/test_data", numTest)
testLabels = samples.loadLabelsFile("data/D2/test_labels", numTest)
featureFunction = dataClassifier.enhancedFeatureExtractorDigit
else:
rawTrainingData = samples.loadDataFile("data/D1/training_data",
numTraining)
trainingLabels = samples.loadLabelsFile("data/D1/training_labels",
numTraining)
rawTestData = samples.loadDataFile("data/D1/test_data", numTest)
testLabels = samples.loadLabelsFile("data/D1/test_labels", numTest)
featureFunction = dataClassifier.basicFeatureExtractorDigit
legalLabels = range(num_classes)
classifier = perceptron1vr.Perceptron1vrClassifier(legalLabels, 3)
# Extract features
print("Extracting features...")
trainingData = map(featureFunction, rawTrainingData)
testData = map(featureFunction, rawTestData)
# Conduct training and testing
print("Training...")
classifier.train(trainingData, trainingLabels, testData, testLabels,
False)
print("Testing...")
guesses = classifier.classify(testData)
correct = [
guesses[i] == testLabels[i] for i in range(len(testLabels))
].count(True)
acc = 100 * correct / (1.0 * (len(testLabels)))
if task == 4:
marks = 0
if (acc > 70):
marks = 3
elif (acc > 60):
marks = 2
elif (acc > 50):
marks = 1
print("Received Marks : " + str(marks) + "/3")
elif task == 6:
marks = 0
print(acc)
if len(testData[0]) <= 5:
if (acc > 85):
marks = 3
elif (acc > 65):
marks = 2
elif (acc > 45):
marks = 1
else:
print("More than permissible features used")
print("Received Marks : " + str(marks) + "/3")
print("--------------------------------------------------------")
def runClassifier(args, options):
featureFunction = args['featureFunction']
classifier = args['classifier']
printImage = args['printImage']
# Load data
trainingFactor = options.training
print "training factor {}".format(trainingFactor)
if options.data == "faces":
TEST_SET_SIZE = 150
numTraining = int(451 * trainingFactor)
print "using {} datapoints out of {} ({}%) for faces".format(
numTraining, 451, 100 * (numTraining / float(451)))
rawTrainingData = samples.loadDataFile("facedata/facedatatrain",
numTraining, FACE_DATUM_WIDTH,
FACE_DATUM_HEIGHT)
trainingLabels = samples.loadLabelsFile("facedata/facedatatrainlabels",
numTraining)
rawValidationData = samples.loadDataFile("facedata/facedatatrain",
TEST_SET_SIZE,
FACE_DATUM_WIDTH,
FACE_DATUM_HEIGHT)
validationLabels = samples.loadLabelsFile(
"facedata/facedatatrainlabels", TEST_SET_SIZE)
rawTestData = samples.loadDataFile("facedata/facedatatest",
TEST_SET_SIZE, FACE_DATUM_WIDTH,
FACE_DATUM_HEIGHT)
testLabels = samples.loadLabelsFile("facedata/facedatatestlabels",
TEST_SET_SIZE)
else:
TEST_SET_SIZE = 1000
numTraining = int(5000 * trainingFactor)
print "using {} datapoints out of {} ({}%) for digits".format(
numTraining, TEST_SET_SIZE, 100 * (numTraining / float(5000)))
rawTrainingData = samples.loadDataFile("digitdata/trainingimages",
numTraining, DIGIT_DATUM_WIDTH,
DIGIT_DATUM_HEIGHT)
trainingLabels = samples.loadLabelsFile("digitdata/traininglabels",
numTraining)
rawValidationData = samples.loadDataFile("digitdata/validationimages",
TEST_SET_SIZE,
DIGIT_DATUM_WIDTH,
DIGIT_DATUM_HEIGHT)
validationLabels = samples.loadLabelsFile("digitdata/validationlabels",
TEST_SET_SIZE)
rawTestData = samples.loadDataFile("digitdata/testimages",
TEST_SET_SIZE, DIGIT_DATUM_WIDTH,
DIGIT_DATUM_HEIGHT)
testLabels = samples.loadLabelsFile("digitdata/testlabels",
TEST_SET_SIZE)
# Extract features
print("Extracting features...")
trainingData = map(featureFunction, rawTrainingData)
validationData = map(featureFunction, rawValidationData)
testData = map(featureFunction, rawTestData)
# Conduct training and testing
automatic = options.run
if automatic:
outcomes = {}
for iterator in range(5):
print("Training...")
if options.data == "faces":
rawTrainingData = samples.loadDataFile(
"facedata/facedatatrain", 451, FACE_DATUM_WIDTH,
FACE_DATUM_HEIGHT)
trainingLabels = samples.loadLabelsFile(
"facedata/facedatatrainlabels", 451)
indices = []
for x in range(numTraining):
indices.append(random.randint(0, 450))
randomTrainingData = []
randomTrainingLabels = []
for index in indices:
randomTrainingData.append(rawTrainingData[index])
randomTrainingLabels.append(trainingLabels[index])
else:
rawTrainingData = samples.loadDataFile(
"digitdata/trainingimages", 5000, DIGIT_DATUM_WIDTH,
DIGIT_DATUM_HEIGHT)
trainingLabels = samples.loadLabelsFile(
"digitdata/traininglabels", 5000)
indices = []
for x in range(numTraining):
indices.append(random.randint(0, 4999))
randomTrainingData = []
randomTrainingLabels = []
for index in indices:
randomTrainingData.append(rawTrainingData[index])
randomTrainingLabels.append(trainingLabels[index])
trainingData = map(featureFunction, randomTrainingData)
start = time.time()
classifier.train(trainingData, randomTrainingLabels,
validationData, validationLabels)
print("Validating...")
guesses = classifier.classify(validationData)
correct = [
guesses[i] == validationLabels[i]
for i in range(len(validationLabels))
].count(True)
print(str(correct),
("correct out of " + str(len(validationLabels)) +
" (%.1f%%).") % (100.0 * correct / len(validationLabels)))
print("Testing...")
guesses = classifier.classify(testData)
correct = [
guesses[i] == testLabels[i] for i in range(len(testLabels))
].count(True)
print(str(correct),
("correct out of " + str(len(testLabels)) + " (%.1f%%).") %
(100.0 * correct / len(testLabels)))
analysis(classifier, guesses, testLabels, testData, rawTestData,
printImage)
interval = time.time() - start
print "Training and testing time: " + str(interval)
outcomes[str(iterator)] = [
"Training and testing time: {}".format(interval),
"accuracy of training: {}%".format(
(100.0 * correct / len(testLabels)))
]
print "outcomes: {}".format(outcomes)
else:
print("Training...")
start = time.time()
classifier.train(trainingData, trainingLabels, validationData,
validationLabels)
interval = time.time() - start
print "Training time: " + str(interval)
print("Validating...")
guesses = classifier.classify(validationData)
correct = [
guesses[i] == validationLabels[i]
for i in range(len(validationLabels))
].count(True)
print(str(correct),
("correct out of " + str(len(validationLabels)) + " (%.1f%%).") %
(100.0 * correct / len(validationLabels)))
print("Testing...")
guesses = classifier.classify(testData)
correct = [
guesses[i] == testLabels[i] for i in range(len(testLabels))
].count(True)
print(str(correct),
("correct out of " + str(len(testLabels)) + " (%.1f%%).") %
(100.0 * correct / len(testLabels)))
analysis(classifier, guesses, testLabels, testData, rawTestData,
printImage)
# do odds ratio computation if specified at command line
if options.odds & (options.classifier != "mostFrequent"):
label1, label2 = options.label1, options.label2
features_odds = classifier.findHighOddsFeatures(label1, label2)
if options.classifier == "naiveBayes" or options.classifier == "nb":
string3 = "=== Features with highest odd ratio of label %d over label %d ===" % (
label1, label2)
else:
string3 = "=== Features for which weight(label %d)-weight(label %d) is biggest ===" % (
label1, label2)
print(string3)
printImage(features_odds)
def runClassifier(args, options):
featureFunction = args['featureFunction']
classifier = args['classifier']
printImage = args['printImage']
# Load data
numTraining = options.training
numTest = options.test
if (options.data == "pacman"):
agentToClone = args.get('agentToClone', None)
trainingData, validationData, testData = MAP_AGENT_TO_PATH_OF_SAVED_GAMES.get(
agentToClone, (None, None, None))
trainingData = trainingData or args.get(
'trainingData',
False) or MAP_AGENT_TO_PATH_OF_SAVED_GAMES['ContestAgent'][0]
validationData = validationData or args.get(
'validationData',
False) or MAP_AGENT_TO_PATH_OF_SAVED_GAMES['ContestAgent'][1]
testData = testData or MAP_AGENT_TO_PATH_OF_SAVED_GAMES[
'ContestAgent'][2]
rawTrainingData, trainingLabels = samples.loadPacmanData(
trainingData, numTraining)
rawValidationData, validationLabels = samples.loadPacmanData(
validationData, numTest)
rawTestData, testLabels = samples.loadPacmanData(testData, numTest)
else:
rawTrainingData = samples.loadDataFile("digitdata/trainingimages",
numTraining, DIGIT_DATUM_WIDTH,
DIGIT_DATUM_HEIGHT)
trainingLabels = samples.loadLabelsFile("digitdata/traininglabels",
numTraining)
rawValidationData = samples.loadDataFile("digitdata/validationimages",
numTest, DIGIT_DATUM_WIDTH,
DIGIT_DATUM_HEIGHT)
validationLabels = samples.loadLabelsFile("digitdata/validationlabels",
numTest)
rawTestData = samples.loadDataFile("digitdata/testimages", numTest,
DIGIT_DATUM_WIDTH,
DIGIT_DATUM_HEIGHT)
testLabels = samples.loadLabelsFile("digitdata/testlabels", numTest)
# Extract features
print "Extracting features..."
trainingData = map(featureFunction, rawTrainingData)
validationData = map(featureFunction, rawValidationData)
testData = map(featureFunction, rawTestData)
# Conduct training and testing
print "Training..."
classifier.train(trainingData, trainingLabels, validationData,
validationLabels)
print "Validating..."
guesses = classifier.classify(validationData)
correct = [
guesses[i] == validationLabels[i] for i in range(len(validationLabels))
].count(True)
print str(correct), ("correct out of " + str(len(validationLabels)) +
" (%.1f%%).") % (100.0 * correct /
len(validationLabels))
print "Testing..."
guesses = classifier.classify(testData)
correct = [guesses[i] == testLabels[i]
for i in range(len(testLabels))].count(True)
print str(correct), ("correct out of " + str(len(testLabels)) +
" (%.1f%%).") % (100.0 * correct / len(testLabels))
analysis(classifier, guesses, testLabels, testData, rawTestData,
printImage)
# do odds ratio computation if specified at command line
if ((options.odds) & (options.classifier == "naiveBayes" or
(options.classifier == "nb"))):
label1, label2 = options.label1, options.label2
features_odds = classifier.findHighOddsFeatures(label1, label2)
if (options.classifier == "naiveBayes" or options.classifier == "nb"):
string3 = "=== Features with highest odd ratio of label %d over label %d ===" % (
label1, label2)
else:
string3 = "=== Features for which weight(label %d)-weight(label %d) is biggest ===" % (
label1, label2)
print string3
printImage(features_odds)
if ((options.weights) & (options.classifier == "perceptron")):
for l in classifier.legalLabels:
features_weights = classifier.findHighWeightFeatures(l)
print("=== Features with high weight for label %d ===" % l)
printImage(features_weights)
def demo_digits(n_images, datapath, labelspath, flag):
start = time.time()
weights_vectors = []
base_path = './TrainDigitsResults/TrainingDigitsResults'
for i in range(0, 10):
load_path = base_path + str(i) + '/100_percent_digit_train.txt'
with open(load_path, 'r') as file:
weights = choose_best_weights(file)
weights_vectors.append(weights)
file.close()
images = samples.loadDataFile(datapath, n_images, 28, 28)
labels = samples.loadLabelsFile(labelspath, n_images)
featureslist = compute_features2(images)
'''
For each image, apply all weight vectors to image
Keep track of each sum in a sums list for the current image,
Choose the sum that:
The highest value
The sum's index in the sums dictionary is the same as the weight vector's index
The weight vector's index is the same as the digit's designated/chosen weights
This index will be what digit you are guessing the image to be
Compare this index/guessed number with the labels[image] value, if they are the same, append True to results
if false, append False to results
'''
results = []
for image in range(len(images)):
sums = []
for weightsvector in range(len(weights_vectors)):
sum = float(0)
for weight in range(len(weights_vectors[weightsvector])):
sum += (weights_vectors[weightsvector][weight] *
featureslist[image][weight])
# Add the sum to the sums dictionary
sums.append(sum)
max = float('-inf')
index = -1
for z in range(10):
if sums[z] > max:
max = sums[z]
index = z
#print(index, end=" --> ")
#print(labels[image])
# Keep track of all guesses/index vs. labels in a tuples list (guess, label)
results.append((index, labels[image]))
correctcount = float(0)
for t in results:
if t[0] == t[1]:
correctcount += float(1)
if flag:
rand = random.randint(0, len(results))
print('Guessed: ' + str(results[rand][0]) + ', Actual: ' +
str(results[rand][1]) + ' (line ' + str(rand) +
' of digit testlabel)')
else:
print(results)
print('Digits accuracy: ' +
str(round((float(correctcount) * 100 / float(len(labels))), 1)) +
'%')
print('Time elapsed: ' + str(round(time.time() - start, 2)) + 's')
def run_digits_n_times():
all_final_stddevs = []
all_final_accs = []
colors = ['r', 'b', 'g', 'orange', 'k', 'c', 'm', 'y', 'grey', 'pink']
fig = plt.figure()
ax1 = fig.add_subplot(111)
ax1.set_title('Face Training Runtimes')
images = samples.loadDataFile('digitdata/trainingimages', 5000, 28, 28)
labels = samples.loadLabelsFile('digitdata/traininglabels', 5000)
# For each percentage
for v in range(1, 11):
sample_percentage = v * 10
acc_list = []
# For all 5 iterations
for j in range(0, 5):
weights_vectors = []
# For each digit 0-9
for y in range(0, 10):
digit = y
images_sample, labels_sample, visited = perceptron.sample_digits(
digit, sample_percentage, images, labels)
# Featureslist = compute_features(functions_list, images_sample, typeflag)
featureslist = perceptron.compute_features2(images_sample)
# Weights = initialize_weights(len(functions_list), 0)
weights = perceptron.initialize_weights(28 * 28, 0)
'''
Run Perceptron / Learn weights
'''
start = time.time()
final_weights = compute_weights(weights, featureslist,
labels_sample)
elapsed = time.time() - start
weights_vectors.append(final_weights)
# Test all weights (weight vectors 0-9), add acc to acc list
acc = demo_digits(weights_vectors)
acc_list.append(acc)
# For percentage sample size v, get the mean accuracy and standard deviation accuracy
mean = statistics.mean(acc_list)
stddev = statistics.stdev(acc_list)
print('Mean accuracy for sample percent-' + str(v) + ' ~ ' + str(mean))
print('STD Dev for sample percent-' + str(v) + ' ~ ' + str(stddev))
all_final_stddevs.append(stddev)
all_final_accs.append(mean)
print(stddev)
ax1.plot([1, 2, 3, 4, 5, 6, 7, 8, 9, 10],
all_final_accs,
color='blue',
marker='.',
linestyle='--')
#ax1.plot([1, 2, 3, 4, 5, 6, 7, 8, 9, 10], all_final_stddevs, color='red',marker='.', linestyle='--')
plt.xticks([1, 2, 3, 4, 5, 6, 7, 8, 9, 10], [
'10%', '20%', '30%', '40%', '50%', '60%', '70%', '80%', '90%', '100%'
])
plt.xlabel('Training Data Sample Size')
plt.ylabel('Standard Deviation of Accuracies')
# plt.legend(loc='lower right', title='Face Train Runtime');
plt.show()
def runClassifier(args, options):
featureFunction = args['featureFunction']
classifier = args['classifier']
printImage = args['printImage']
# Load data
numTraining = options.training
numTest = options.test
trainingTime = []
if (options.data == "faces"):
rawTrainingData = samples.loadDataFile("facedata/facedatatrain",
numTraining, FACE_DATUM_WIDTH,
FACE_DATUM_HEIGHT)
trainingLabels = samples.loadLabelsFile("facedata/facedatatrainlabels",
numTraining)
rawValidationData = samples.loadDataFile("facedata/facedatatrain",
numTest, FACE_DATUM_WIDTH,
FACE_DATUM_HEIGHT)
validationLabels = samples.loadLabelsFile(
"facedata/facedatatrainlabels", numTest)
rawTestData = samples.loadDataFile("facedata/facedatatest", numTest,
FACE_DATUM_WIDTH, FACE_DATUM_HEIGHT)
testLabels = samples.loadLabelsFile("facedata/facedatatestlabels",
numTest)
else:
rawTrainingData = samples.loadDataFile("digitdata/trainingimages",
numTraining, DIGIT_DATUM_WIDTH,
DIGIT_DATUM_HEIGHT)
trainingLabels = samples.loadLabelsFile("digitdata/traininglabels",
numTraining)
rawValidationData = samples.loadDataFile("digitdata/validationimages",
numTest, DIGIT_DATUM_WIDTH,
DIGIT_DATUM_HEIGHT)
validationLabels = samples.loadLabelsFile("digitdata/validationlabels",
numTest)
rawTestData = samples.loadDataFile("digitdata/testimages", numTest,
DIGIT_DATUM_WIDTH,
DIGIT_DATUM_HEIGHT)
testLabels = samples.loadLabelsFile("digitdata/testlabels", numTest)
f = open("training_time", "a")
f.write("Classifier: {}\n".format(options.classifier))
f.write("Data: {}\n".format(options.data))
f.write("{} training\n".format(options.training))
f.write("{} testing\n".format(options.test))
# Extract features
print("Extracting features...")
trainingData = map(featureFunction, rawTrainingData)
validationData = map(featureFunction, rawValidationData)
testData = map(featureFunction, rawTestData)
# Conduct training and Testing
print("Training...")
start_stamp = time.time()
classifier.train(trainingData, trainingLabels, validationData,
validationLabels)
end_stamp = time.time()
trainingTime.append(str(end_stamp - start_stamp))
print("Validating...")
guesses = classifier.classify(validationData)
correct = [
guesses[i] == validationLabels[i] for i in range(len(validationLabels))
].count(True)
print str(correct), ("correct out of " + str(len(validationLabels)) +
" (%.1f%%).") % (100.0 * correct /
len(validationLabels))
print("Testing...")
guesses = classifier.classify(testData)
correct = [guesses[i] == testLabels[i]
for i in range(len(testLabels))].count(True)
print str(correct), ("correct out of " + str(len(testLabels)) +
" (%.1f%%).") % (100.0 * correct / len(testLabels))
analysis(classifier, guesses, testLabels, testData, rawTestData,
printImage)
# print str(guesses) ###############################################################################################################################
accuracy = str(100.0 * correct / len(testLabels))
# do odds ratio computation if specified at command line
if ((options.odds) & (options.classifier == "naiveBayes" or
(options.classifier == "nb"))):
label1, label2 = options.label1, options.label2
features_odds = classifier.findHighOddsFeatures(label1, label2)
if (options.classifier == "naiveBayes" or options.classifier == "nb"):
string3 = "=== Features with highest odd ratio of label %d over label %d ===" % (
label1, label2)
else:
string3 = "=== Features for which weight(label %d)-weight(label %d) is biggest ===" % (
label1, label2)
#print (string3)
printImage(features_odds)
if ((options.weights) & (options.classifier == "perceptron")):
for l in classifier.legalLabels:
features_weights = classifier.findHighWeightFeatures(l)
print("=== Features with high weight for label %d ===" % l)
printImage(features_weights)
print("training time", trainingTime)
f.write("{}\n".format(trainingTime))
f.write("{}\n".format(accuracy))
f.write("\n")
f.close()
def runClassifier(args, options):
classifier = args['classifier']
# Load data
if options.data == "pacman":
agentToClone = args.get('agentToClone', None)
trainingData, validationData, testData = MAP_AGENT_TO_PATH_OF_SAVED_GAMES.get(
agentToClone, (None, None, None))
trainingData = trainingData or args.get(
'trainingData',
False) or MAP_AGENT_TO_PATH_OF_SAVED_GAMES['ContestAgent'][0]
validationData = validationData or args.get(
'validationData',
False) or MAP_AGENT_TO_PATH_OF_SAVED_GAMES['ContestAgent'][1]
testData = testData or MAP_AGENT_TO_PATH_OF_SAVED_GAMES[
'ContestAgent'][2]
trainingData, trainingLabels = samples.loadPacmanData(
trainingData, options.training)
validationData, validationLabels = samples.loadPacmanData(
validationData, options.validation)
testData, testLabels = samples.loadPacmanData(testData, None)
elif options.data == "digits":
if options.training is None:
options.training = 2000
if options.validation is None:
options.validation = 1000
numTest = 1000
trainingData = samples.loadDataFile("digitdata/trainingimages",
options.training,
DIGIT_DATUM_WIDTH,
DIGIT_DATUM_HEIGHT)
trainingLabels = samples.loadLabelsFile("digitdata/traininglabels",
options.training)
validationData = samples.loadDataFile("digitdata/validationimages",
options.validation,
DIGIT_DATUM_WIDTH,
DIGIT_DATUM_HEIGHT)
validationLabels = samples.loadLabelsFile("digitdata/validationlabels",
options.validation)
testData = samples.loadDataFile("digitdata/testimages", numTest,
DIGIT_DATUM_WIDTH, DIGIT_DATUM_HEIGHT)
testLabels = samples.loadLabelsFile("digitdata/testlabels", numTest)
else:
raise ValueError('unrecognized dataset %r' % options.data)
# Conduct training and testing
print "Training..."
classifier.train(trainingData, trainingLabels, validationData,
validationLabels)
print "Testing..."
guesses = classifier.classify(testData)
correct = [guesses[i] == testLabels[i]
for i in range(len(testLabels))].count(True)
print str(correct), ("correct out of " + str(len(testLabels)) +
" (%.1f%%).") % (100.0 * correct / len(testLabels))
def runClassifier(args, options):
#print 'args: ', args
#print 'options', options
featureFunction = args['featureFunction']
classifier = args['classifier']
printImage = args['printImage']
# Load data
numTraining = options.training
numTest = options.test
if(options.data=="faces"):
rawTrainingData = samples.loadDataFile("facedata/facedatatrain", numTraining,FACE_DATUM_WIDTH,FACE_DATUM_HEIGHT)
trainingLabels = samples.loadLabelsFile("facedata/facedatatrainlabels", numTraining)
rawValidationData = samples.loadDataFile("facedata/facedatatrain", numTest,FACE_DATUM_WIDTH,FACE_DATUM_HEIGHT)
validationLabels = samples.loadLabelsFile("facedata/facedatatrainlabels", numTest)
rawTestData = samples.loadDataFile("facedata/facedatatest", numTest,FACE_DATUM_WIDTH,FACE_DATUM_HEIGHT)
testLabels = samples.loadLabelsFile("facedata/facedatatestlabels", numTest)
else:
rawTrainingData = samples.loadDataFile("digitdata/trainingimages", numTraining,DIGIT_DATUM_WIDTH,DIGIT_DATUM_HEIGHT)
trainingLabels = samples.loadLabelsFile("digitdata/traininglabels", numTraining)
rawValidationData = samples.loadDataFile("digitdata/validationimages", numTest,DIGIT_DATUM_WIDTH,DIGIT_DATUM_HEIGHT)
validationLabels = samples.loadLabelsFile("digitdata/validationlabels", numTest)
rawTestData = samples.loadDataFile("digitdata/testimages", numTest,DIGIT_DATUM_WIDTH,DIGIT_DATUM_HEIGHT)
testLabels = samples.loadLabelsFile("digitdata/testlabels", numTest)
# Extract features
#print "Extracting features..."
#print '#######type of rawTrainingData is', rawTrainingData.__class__ # list of Datum
#print '#######type of rawTrainingData[0] is', rawTrainingData[0].__class__ # Datum
trainingData = map(featureFunction, rawTrainingData)
#print '#######type of trainingData is', trainingData.__class__ # list of Counter
#print '#######type of trainingData[0] is', trainingData[0].__class__ # Counter
validationData = map(featureFunction, rawValidationData)
testData = map(featureFunction, rawTestData)
# Conduct training and testing
print "Training..."
classifier.train(trainingData, trainingLabels, validationData, validationLabels)
print "Validating..."
guesses = classifier.classify(validationData)
print 'length of guesses is %d' % len(guesses)
correct = [guesses[i] == validationLabels[i] for i in range(len(validationLabels))].count(True)
print str(correct), ("correct out of " + str(len(validationLabels)) + " (%.1f%%).") % (100.0 * correct / len(validationLabels))
print "Testing..."
guesses = classifier.classify(testData)
correct = [guesses[i] == testLabels[i] for i in range(len(testLabels))].count(True)
print str(correct), ("correct out of " + str(len(testLabels)) + " (%.1f%%).") % (100.0 * correct / len(testLabels))
analysis(classifier, guesses, testLabels, testData, rawTestData, printImage)
# do odds ratio computation if specified at command line
if((options.odds) & (options.classifier == "naiveBayes" or (options.classifier == "nb")) ):
label1, label2 = options.label1, options.label2
features_odds = classifier.findHighOddsFeatures(label1,label2)
if(options.classifier == "naiveBayes" or options.classifier == "nb"):
string3 = "=== Features with highest odd ratio of label %d over label %d ===" % (label1, label2)
else:
string3 = "=== Features for which weight(label %d)-weight(label %d) is biggest ===" % (label1, label2)
print string3
printImage(features_odds)
if((options.weights) & (options.classifier == "perceptron")):
for l in classifier.legalLabels:
features_weights = classifier.findHighWeightFeatures(l)
print ("=== Features with high weight for label %d ==="%l)
printImage(features_weights)
def runClassifier():
global TK_ROOT, SP_CANVAS, LOG_X, LOG_Y
# Set up variables according to the command line inputs
featureFunction = basicFeatureExtractorDigit
legalLabels = range(10) # number of labels
# Select classifier
classifier = perceptron.PerceptronClassifier(legalLabels)
# Load data
numTraining = 1
loadImage()
rawTrainingData = samples.loadDataFile("digitdata/trainingimages",
numTraining, DIGIT_DATUM_WIDTH,
DIGIT_DATUM_HEIGHT, 'train',
SP_CANVAS)
trainingLabels = samples.loadLabelsFile("digitdata/traininglabels",
numTraining)
rawTestData = samples.loadDataFile("digitdata/testingimages",
TEST_SET_SIZE, DIGIT_DATUM_WIDTH,
DIGIT_DATUM_HEIGHT, 'test', SP_CANVAS)
testLabels = samples.loadLabelsFile("digitdata/testlabels", TEST_SET_SIZE)
# Extract features
print rawTestData
trainingData = map(basicFeatureExtractorDigit, rawTrainingData)
print "cp3"
testData = map(basicFeatureExtractorDigit, rawTestData)
# Conduct auto training
SP_CANVAS.create_text(LOG_X,
LOG_Y,
text="Auto Training...",
anchor=NW,
font=tkFont.Font(size=-14))
LOG_Y += 15
classifier.train(trainingData, trainingLabels, SP_CANVAS)
# Auto Testing
# print "Validating..."
# guesses = classifier.classify(validationData)
# correct = [guesses[i] == validationLabels[i] for i in range(len(validationLabels))].count(True)
# print str(correct), ("correct out of " + str(len(validationLabels)) + " (%.1f%%).") % (100.0 * correct / len(validationLabels))
# User Input Testing
SP_CANVAS.create_text(LOG_X,
LOG_Y,
text="Recognizing...",
anchor=NW,
font=tkFont.Font(size=-14))
LOG_Y += 15
guesses = classifier.classify(testData, SP_CANVAS, "usr")
# Completion Notify
SP_CANVAS.create_text(LOG_X,
LOG_Y + 30,
text="Completed...",
anchor=NW,
font=tkFont.Font(size=-14))
LOG_Y += 15
def runClassifier(args, options):
featureFunction = args['featureFunction']
classifier = args['classifier']
printImage = args['printImage']
# Load the data for testing, training and validation
if(options.random):
numberOfTestPoints = 150 if options.data=="faces" else 1000
numberOfValidationPoints = 301 if options.data=="faces" else 1000
totalTrainData = 451 if options.data=="faces" else 5000
numValidation = numberOfValidationPoints
numTest = numberOfTestPoints
numTraining = totalTrainData
if(options.data=="faces"):
rawTestData = samples.loadDataFile("facedata/facedatatest", numTest,FACE_DATUM_WIDTH,FACE_DATUM_HEIGHT)
testLabels = samples.loadLabelsFile("facedata/facedatatestlabels", numTest)
rawValidationData = samples.loadDataFile("facedata/facedatatrain", numValidation,FACE_DATUM_WIDTH,FACE_DATUM_HEIGHT)
validationLabels = samples.loadLabelsFile("facedata/facedatatrainlabels", numValidation)
rawTrainingData = samples.loadDataFile("facedata/facedatatrain", numTraining,FACE_DATUM_WIDTH,FACE_DATUM_HEIGHT)
trainingLabels = samples.loadLabelsFile("facedata/facedatatrainlabels", numTraining)
else:
rawTestData = samples.loadDataFile("digitdata/testimages", numTest,DIGIT_DATUM_WIDTH,DIGIT_DATUM_HEIGHT)
testLabels = samples.loadLabelsFile("digitdata/testlabels", numTest)
rawValidationData = samples.loadDataFile("digitdata/validationimages", numValidation,DIGIT_DATUM_WIDTH,DIGIT_DATUM_HEIGHT)
validationLabels = samples.loadLabelsFile("digitdata/validationlabels", numValidation)
rawTrainingData = samples.loadDataFile("digitdata/trainingimages", numTraining,DIGIT_DATUM_WIDTH,DIGIT_DATUM_HEIGHT)
trainingLabels = samples.loadLabelsFile("digitdata/traininglabels", numTraining)
print ("Extracting features...")
trainingData = []
validationData = []
testData = []
for datum1 in rawTestData:
k = featureFunction(datum1)
testData.append(k)
for datum1 in rawValidationData:
k = featureFunction(datum1)
validationData.append(k)
for datum1 in rawTrainingData:
k = featureFunction(datum1)
trainingData.append(k)
for percent in range(1,11):
accuracy = []
times = []
print("\n")
for runCount in range(0,5):
# Extract features
print("======================================\n")
print ("("+str(runCount+1)+")" + " Extracting random " + str((percent * 10)) + "% of the training data...")
numSubTraining = int((percent / 10.0) * totalTrainData)
indexes = random.sample(range(0, totalTrainData), numSubTraining)
subTrainingData = []
subTrainingLabels = []
for indx in indexes:
subTrainingData.append(trainingData[indx])
subTrainingLabels.append(trainingLabels[indx])
# Conduct training and testing
start = time.time()
print ("("+str(runCount + 1)+")", "Training on", numSubTraining, "data points...")
classifier.train(subTrainingData, subTrainingLabels, validationData, validationLabels)
end = time.time()
elapsed = end - start
print ("("+str(runCount + 1)+")" + " Training completed in %0.4f second(s)" % elapsed)
times.append(elapsed)
# Validation
print ("("+str(runCount+1)+")", "Validating...")
guesses = classifier.classify(validationData)
correct = [guesses[i] == validationLabels[i] for i in range(len(validationLabels))].count(True)
print ("("+str(runCount + 1)+") " + str(correct), ("correct out of " + str(len(validationLabels)) + " (%.1f%%).") % (100.0 * correct / len(validationLabels)))
# Testing
print ("("+str(runCount+1)+")", "Testing...")
guesses = classifier.classify(testData)
correct = [guesses[i] == testLabels[i] for i in range(len(testLabels))].count(True)
print ("("+str(runCount + 1)+") " + str(correct), ("correct out of " + str(len(testLabels)) + " (%.1f%%).") % (100.0 * correct / len(testLabels)) + "\n")
accuracy.append(100.0 * correct / len(testLabels))
averageAccuracy = 0
avg_time = 0
for q in range(0, len(accuracy)):
averageAccuracy += accuracy[q]
avg_time += times[q]
averageAccuracy = averageAccuracy/len(accuracy)
avg_time = avg_time/len(times)
print("=================\n")
print ("Average training time for", numSubTraining, "data points: %0.4f" % avg_time)
print ("Average accuracy of " + str(percent * 10) + ("% data training: "), str(averageAccuracy))
std_dev = 0
for a in accuracy:
temp = a - averageAccuracy
std_dev += (temp*temp)
std_dev = std_dev / (len(accuracy) - 1)
std_dev = math.sqrt(std_dev)
print ("Standard deviation of accuracy: %0.4f" % std_dev)
print
sys.exit(1)
else:
numTraining = options.training
numTest = options.test
classifier.extra = True
if(options.data=="faces"):
rawTrainingData = samples.loadDataFile("facedata/facedatatrain", numTraining,FACE_DATUM_WIDTH,FACE_DATUM_HEIGHT)
trainingLabels = samples.loadLabelsFile("facedata/facedatatrainlabels", numTraining)
rawValidationData = samples.loadDataFile("facedata/facedatatrain", numTest,FACE_DATUM_WIDTH,FACE_DATUM_HEIGHT)
validationLabels = samples.loadLabelsFile("facedata/facedatatrainlabels", numTest)
rawTestData = samples.loadDataFile("facedata/facedatatest", numTest,FACE_DATUM_WIDTH,FACE_DATUM_HEIGHT)
testLabels = samples.loadLabelsFile("facedata/facedatatestlabels", numTest)
else:
rawTrainingData = samples.loadDataFile("digitdata/trainingimages", numTraining,DIGIT_DATUM_WIDTH,DIGIT_DATUM_HEIGHT)
trainingLabels = samples.loadLabelsFile("digitdata/traininglabels", numTraining)
rawValidationData = samples.loadDataFile("digitdata/validationimages", numTest,DIGIT_DATUM_WIDTH,DIGIT_DATUM_HEIGHT)
validationLabels = samples.loadLabelsFile("digitdata/validationlabels", numTest)
rawTestData = samples.loadDataFile("digitdata/testimages", numTest,DIGIT_DATUM_WIDTH,DIGIT_DATUM_HEIGHT)
testLabels = samples.loadLabelsFile("digitdata/testlabels", numTest)
# Extract features
print ("Extracting features...")
trainingData = []
validationData = []
testData = []
for datum1 in rawTestData:
k = featureFunction(datum1)
testData.append(k)
# trainingData = map(featureFunction, rawValidationData)
for datum1 in rawValidationData:
k = featureFunction(datum1)
validationData.append(k)
for datum1 in rawTrainingData:
k = featureFunction(datum1)
trainingData.append(k)
# Conduct training and testing
print ("Training...")
classifier.train(trainingData, trainingLabels, validationData, validationLabels)
print ("Validating...")
guesses = classifier.classify(validationData)
correct = [guesses[i] == validationLabels[i] for i in range(len(validationLabels))].count(True)
print (str(correct), ("correct out of " + str(len(validationLabels)) + " (%.1f%%).") % (100.0 * correct / len(validationLabels)))
print ("Testing...")
guesses = classifier.classify(testData)
correct = [guesses[i] == testLabels[i] for i in range(len(testLabels))].count(True)
print (str(correct), ("correct out of " + str(len(testLabels)) + " (%.1f%%).") % (100.0 * correct / len(testLabels)))
analysis(classifier, guesses, testLabels, testData, rawTestData, printImage)
# HighOddsFeatures
if((options.odds) & (options.classifier == "naiveBayes" or (options.classifier == "nb")) ):
label1, label2 = options.label1, options.label2
featOdds = classifier.findHighOddsFeatures(label1,label2)
if(options.classifier == "naiveBayes" or options.classifier == "nb"):
feats = "=== Features with highest odd ratio of label %d over label %d ===" % (label1, label2)
else:
feats = "=== Features for which weight (label %d) - weight (label %d) is largest ===" % (label1, label2)
print (feats)
printImage(featOdds)
if((options.weights) & (options.classifier == "perceptron")):
for l in classifier.legalLabels:
featWeights = classifier.findHighWeightFeatures(l)
print ("=== Features with high weight for label %d ===" % l)
printImage(featWeights)
def runClassifier(args, options):
featureFunction = args['featureFunction']
classifier = args['classifier']
printImage = args['printImage']
# Load data
numTraining = options.training
numTest = options.test
if (options.data == "faces"):
rawTrainingData = samples.loadDataFile("facedata/facedatatrain", numTraining, FACE_DATUM_WIDTH,
FACE_DATUM_HEIGHT)
trainingLabels = samples.loadLabelsFile("facedata/facedatatrainlabels", numTraining)
rawValidationData = samples.loadDataFile("facedata/facedatatrain", numTest, FACE_DATUM_WIDTH, FACE_DATUM_HEIGHT)
validationLabels = samples.loadLabelsFile("facedata/facedatatrainlabels", numTest)
rawTestData = samples.loadDataFile("facedata/facedatatest", numTest, FACE_DATUM_WIDTH, FACE_DATUM_HEIGHT)
testLabels = samples.loadLabelsFile("facedata/facedatatestlabels", numTest)
else:
rawTrainingData = samples.loadDataFile("digitdata/trainingimages", numTraining, DIGIT_DATUM_WIDTH,
DIGIT_DATUM_HEIGHT)
trainingLabels = samples.loadLabelsFile("digitdata/traininglabels", numTraining)
rawValidationData = samples.loadDataFile("digitdata/validationimages", numTest, DIGIT_DATUM_WIDTH,
DIGIT_DATUM_HEIGHT)
validationLabels = samples.loadLabelsFile("digitdata/validationlabels", numTest)
rawTestData = samples.loadDataFile("digitdata/testimages", numTest, DIGIT_DATUM_WIDTH, DIGIT_DATUM_HEIGHT)
testLabels = samples.loadLabelsFile("digitdata/testlabels", numTest)
# Extract features
print("Extracting features...")
# randomly choose 10% percentage of the dataset from the training set
percentage = 0.1
rand_sample = [i for i in sorted(random.sample(range(numTraining), int(numTraining * percentage)))]
sample_rawTrainingData = [rawTrainingData[i] for i in rand_sample]
sample_trainingLabels = [trainingLabels[i] for i in rand_sample]
trainingData = list(map(featureFunction, sample_rawTrainingData))
validationData = list(map(featureFunction, rawValidationData))
testData = list(map(featureFunction, rawTestData))
# Conduct training and testing
print("Training...")
start = timeit.default_timer()
classifier.train(trainingData, sample_trainingLabels, validationData, validationLabels)
stop = timeit.default_timer()
print("training time is: ", stop - start)
print("Validating...")
guesses = classifier.classify(validationData)
correct = [guesses[i] == validationLabels[i] for i in range(len(validationLabels))].count(True)
print(str(correct),
("correct out of " + str(len(validationLabels)) + " (%.1f%%).") % (100.0 * correct / len(validationLabels)))
print("Testing...")
guesses = classifier.classify(testData)
correct = [guesses[i] == testLabels[i] for i in range(len(testLabels))].count(True)
print(str(correct), ("correct out of " + str(len(testLabels)) + " (%.1f%%).") % (100.0 * correct / len(testLabels)))
analysis(classifier, guesses, testLabels, testData, rawTestData, printImage)
# do odds ratio computation if specified at command line
if ((options.odds) & (options.classifier == "naiveBayes" or (options.classifier == "nb"))):
label1, label2 = options.label1, options.label2
features_odds = classifier.findHighOddsFeatures(label1, label2)
if (options.classifier == "naiveBayes" or options.classifier == "nb"):
string3 = "=== Features with highest odd ratio of label %d over label %d ===" % (label1, label2)
else:
string3 = "=== Features for which weight(label %d)-weight(label %d) is biggest ===" % (label1, label2)
print(string3)
printImage(features_odds)
if ((options.weights) & (options.classifier == "perceptron")):
for l in classifier.legalLabels:
features_weights = classifier.findHighWeightFeatures(l)
print("=== Features with high weight for label %d ===" % l)
printImage(features_weights)
def runClassifier(args, options):
featureFunction = args['featureFunction']
classifier = args['classifier']
printImage = args['printImage']
# Load data
numTraining = options.training
numTest = options.test
if(options.data=="pacman"):
agentToClone = args.get('agentToClone', None)
trainingData, validationData, testData = MAP_AGENT_TO_PATH_OF_SAVED_GAMES.get(agentToClone, (None, None, None))
trainingData = trainingData or args.get('trainingData', False) or MAP_AGENT_TO_PATH_OF_SAVED_GAMES['ContestAgent'][0]
validationData = validationData or args.get('validationData', False) or MAP_AGENT_TO_PATH_OF_SAVED_GAMES['ContestAgent'][1]
testData = testData or MAP_AGENT_TO_PATH_OF_SAVED_GAMES['ContestAgent'][2]
rawTrainingData, trainingLabels = samples.loadPacmanData(trainingData, numTraining)
rawValidationData, validationLabels = samples.loadPacmanData(validationData, numTest)
rawTestData, testLabels = samples.loadPacmanData(testData, numTest)
else:
rawTrainingData = samples.loadDataFile("digitdata/trainingimages", numTraining,DIGIT_DATUM_WIDTH,DIGIT_DATUM_HEIGHT)
trainingLabels = samples.loadLabelsFile("digitdata/traininglabels", numTraining)
rawValidationData = samples.loadDataFile("digitdata/validationimages", numTest,DIGIT_DATUM_WIDTH,DIGIT_DATUM_HEIGHT)
validationLabels = samples.loadLabelsFile("digitdata/validationlabels", numTest)
rawTestData = samples.loadDataFile("digitdata/testimages", numTest,DIGIT_DATUM_WIDTH,DIGIT_DATUM_HEIGHT)
testLabels = samples.loadLabelsFile("digitdata/testlabels", numTest)
# Extract features
print "Extracting features..."
trainingData = map(featureFunction, rawTrainingData)
validationData = map(featureFunction, rawValidationData)
testData = map(featureFunction, rawTestData)
# Conduct training and testing
print "Training..."
classifier.train(trainingData, trainingLabels, validationData, validationLabels)
print "Validating..."
guesses = classifier.classify(validationData)
correct = [guesses[i] == validationLabels[i] for i in range(len(validationLabels))].count(True)
print str(correct), ("correct out of " + str(len(validationLabels)) + " (%.1f%%).") % (100.0 * correct / len(validationLabels))
print "Testing..."
guesses = classifier.classify(testData)
correct = [guesses[i] == testLabels[i] for i in range(len(testLabels))].count(True)
print str(correct), ("correct out of " + str(len(testLabels)) + " (%.1f%%).") % (100.0 * correct / len(testLabels))
analysis(classifier, guesses, testLabels, testData, rawTestData, printImage)
# do odds ratio computation if specified at command line
if((options.odds) & (options.classifier == "naiveBayes" or (options.classifier == "nb")) ):
label1, label2 = options.label1, options.label2
features_odds = classifier.findHighOddsFeatures(label1,label2)
if(options.classifier == "naiveBayes" or options.classifier == "nb"):
string3 = "=== Features with highest odd ratio of label %d over label %d ===" % (label1, label2)
else:
string3 = "=== Features for which weight(label %d)-weight(label %d) is biggest ===" % (label1, label2)
print string3
printImage(features_odds)
if((options.weights) & (options.classifier == "perceptron")):
for l in classifier.legalLabels:
features_weights = classifier.findHighWeightFeatures(l)
print ("=== Features with high weight for label %d ==="%l)
printImage(features_weights)
def runClassifier(args, options):
featureFunction = args['featureFunction']
classifier = args['classifier']
printImage = args['printImage']
# Load data
numTraining = options.training
# Extract features
print "Extracting features..."
if options.data=="faces":
rawTrainingData = samples.loadDataFile("facedata/facedatatrain", numTraining,FACE_DATUM_WIDTH,FACE_DATUM_HEIGHT)
trainingLabels = samples.loadLabelsFile("facedata/facedatatrainlabels", numTraining)
rawValidationData = samples.loadDataFile("facedata/facedatavalidation", TEST_SET_SIZE,FACE_DATUM_WIDTH,FACE_DATUM_HEIGHT)
validationLabels = samples.loadLabelsFile("facedata/facedatavalidationlabels", TEST_SET_SIZE)
rawTestData = samples.loadDataFile("facedata/facedatatest", TEST_SET_SIZE,FACE_DATUM_WIDTH,FACE_DATUM_HEIGHT)
testLabels = samples.loadLabelsFile("facedata/facedatatestlabels", TEST_SET_SIZE)
else:
rawTrainingData = samples.loadDataFile("digitdata/trainingimages", numTraining,DIGIT_DATUM_WIDTH,DIGIT_DATUM_HEIGHT)
trainingLabels = samples.loadLabelsFile("digitdata/traininglabels", numTraining)
rawValidationData = samples.loadDataFile("digitdata/validationimages", TEST_SET_SIZE,DIGIT_DATUM_WIDTH,DIGIT_DATUM_HEIGHT)
validationLabels = samples.loadLabelsFile("digitdata/validationlabels", TEST_SET_SIZE)
rawTestData = samples.loadDataFile("digitdata/testimages", TEST_SET_SIZE,DIGIT_DATUM_WIDTH,DIGIT_DATUM_HEIGHT)
testLabels = samples.loadLabelsFile("digitdata/testlabels", TEST_SET_SIZE)
if options.classifier == "GDA" or options.classifier == "LR":
import os.path
if os.path.isfile(options.data + '_' + str(numTraining) + '_pca.np'):
f = open(options.data + '_' + str(numTraining) + '_pca.np', 'rb')
principleComponents, trainingData, validationData, testData = cPickle.load(f)
f.close()
else:
if options.data == "faces":
dimension = 13
principleComponents = getPrincipleComponents(map(featureFunction, samples.loadDataFile("facedata/facedatatrain",451,FACE_DATUM_WIDTH,FACE_DATUM_HEIGHT)), dimension)
trainingData = np.dot(basicFeatureDataToNumpyArray(map(featureFunction, rawTrainingData)), principleComponents)
validationData = np.dot(basicFeatureDataToNumpyArray(map(featureFunction, rawValidationData)), principleComponents)
testData = np.dot(basicFeatureDataToNumpyArray(map(featureFunction, rawTestData)), principleComponents)
else:
dimension = 13
principleComponents = getPrincipleComponents(map(featureFunction, samples.loadDataFile("digitdata/trainingimages",5000,DIGIT_DATUM_WIDTH,DIGIT_DATUM_HEIGHT)), dimension)
trainingData = np.dot(basicFeatureDataToNumpyArray(map(featureFunction, rawTrainingData)), principleComponents)
validationData = np.dot(basicFeatureDataToNumpyArray(map(featureFunction, rawValidationData)), principleComponents)
testData = np.dot(basicFeatureDataToNumpyArray(map(featureFunction, rawTestData)), principleComponents)
f = open(options.data + '_' + str(numTraining) + '_pca.np', 'wb')
cPickle.dump((principleComponents, trainingData, validationData, testData), f)
f.close()
elif options.classifier == "GPC":
trainingData = basicFeatureDataToNumpyArray(map(featureFunction, rawTrainingData))
validationData = basicFeatureDataToNumpyArray(map(featureFunction, rawValidationData))
testData = basicFeatureDataToNumpyArray(map(featureFunction, rawTestData))
else:
trainingData = map(featureFunction, rawTrainingData)
validationData = map(featureFunction, rawValidationData)
testData = map(featureFunction, rawTestData)
# Conduct training and testing
print "Training..."
classifier.train(trainingData, trainingLabels, validationData, validationLabels)
print "Validating..."
guesses = classifier.classify(validationData)
correct = [guesses[i] == validationLabels[i] for i in range(len(validationLabels))].count(True)
print str(correct), ("correct out of " + str(len(validationLabels)) + " (%.1f%%).") % (100.0 * correct / len(validationLabels))
print "Testing..."
guesses = classifier.classify(testData)
correct = [guesses[i] == testLabels[i] for i in range(len(testLabels))].count(True)
print str(correct), ("correct out of " + str(len(testLabels)) + " (%.1f%%).") % (100.0 * correct / len(testLabels))
analysis(classifier, guesses, testLabels, testData, rawTestData, printImage)
