def testClassic():
m = classifiers.KNNClassifier(tr_images_S, tr_labels_O, lambda x: analysis.mean(x), k=10)
m.evaluate(test_images_S, test_labels_O, False)
h = classifiers.KNNClassifier(tr_images_O, tr_labels_O, lambda x: analysis.histogram(x), k=10)
h.evaluate(test_images_O, test_labels_O, False)
hnn = hogNN()
hnn.evaluate(test_images_O, test_labels_O, False)
mc = classifiers.EnsembleClassifier(confidence_mode = "average")
mc.add(hnn, 0)
mc.add(m, 1)
mc.evaluateMixedImageSets(test_images_O + test_images_S, test_labels_O, False)
mc.add(hist(), 0)
mc.evaluateMixedImageSets(test_images_O + test_images_S, test_labels_O, False)
def hogNearestNeighbour():
knn = classifiers.KNNClassifier(tr_images_O, tr_labels_O, lambda x: hog(resize(x, imagesize, mode = "constant", anti_aliasing=True),
block_norm = "L2-Hys", orientations = 8,
pixels_per_cell = (14, 14), cells_per_block = (1, 1),
visualize = False, multichannel = True), k=5)
return knn
def readCommand( argv ):
"Processes the command used to run from the command line."
global ClassifierText
from optparse import OptionParser
parser = OptionParser(USAGE_STRING)
parser.add_option('-c', '--classifier', help=default('The type of classifier'), choices=['knn', 'perceptron', 'svm', 'lr', 'best', 'none'], default='none')
parser.add_option('-d', '--data', help=default('Dataset to use'), choices=['digits'], default='digits')
parser.add_option('-t', '--training', help=default('The size of the training set'), default=TRAIN_SET_SIZE, type="int")
parser.add_option('-i', '--iterations', help=default("Maximum iterations to run training"), default=50, type="int")
parser.add_option('-s', '--test', help=default("Amount of test data to use"), default=TEST_SET_SIZE, type="int")
parser.add_option('-n', '--neighbors', help=default("Number of clusters(for KMeans) or nearest neighbors(for KNN)"), default=5, type="int")
parser.add_option('-v', '--visualize', help=default('Whether to visualize some results'), action='store_true')
options, otherjunk = parser.parse_args(argv)
if len(otherjunk) != 0: raise Exception('Command line input not understood: ' + str(otherjunk))
args = {}
# Set up variables according to the command line input.
print "Doing classification"
print "--------------------"
print "data:\t\t" + options.data
if options.data == 'digits':
print "classifier:\t\t" + options.classifier
else:
print "classifier:\t\t" + options.data + " classifier"
print "training set size:\t" + str(options.training)
featureExtractor = BasicFeatureExtractorDigit()
legalLabels = range(10)
if options.training <= 0:
print "Training set size should be a positive integer (you provided: %d)" % options.training
print USAGE_STRING
sys.exit(2)
if options.classifier == "perceptron":
classifier = classifiers.PerceptronClassifier(legalLabels, options.iterations)
elif options.classifier == "svm":
classifier = classifiers.SVMClassifier(legalLabels)
elif options.classifier == "knn":
classifier = classifiers.KNNClassifier(legalLabels, options.neighbors)
elif options.classifier == 'lr':
classifier = classifiers.LinearRegressionClassifier(legalLabels)
elif options.classifier == 'best':
classifier = classifiers.BestClassifier(legalLabels)
else:
print "Do not use any classifier:", options.classifier
classifier = None
args['classifier'] = classifier
args['featureExtractor'] = featureExtractor
return args, options
def q1(grades):
print "\n===== Grading Q1 ===== "
grades[1] = 0
print "1. checking python dataClassifier.py -d digits -c knn -t 5000 -s 1000 -f basic -n 5"
try:
legalLabels = range(10)
classifier = classifiers.KNNClassifier(legalLabels, 5)
accDigits = getAccuracy(digitData, classifier) # our solution: 91.0%
accThres = [50, 70, 80, 87]
score = [accDigits >= t for t in accThres].count(True)
grades[1] = score
checkSklearnPackage(grades, 1)
except:
display("An exception was raised:\n%s" % getExceptionTraceBack())
def hist():
knn = classifiers.KNNClassifier(tr_images_O, tr_labels_O, lambda x: analysis.histogram(x), k=10)
return knn
def mean(imagesize = imagesize):
knn = classifiers.KNNClassifier(tr_images_O, tr_labels_O, lambda x: analysis.mean(x), k=10)
return knn
def readCommand( argv ):
"Processes the command used to run from the command line."
global ClassifierText
from optparse import OptionParser
parser = OptionParser(USAGE_STRING)
parser.add_option('-c', '--classifier', help=default('The type of classifier'), choices=['knn', 'perceptron', 'svm', 'lr', 'none'], default='none')
parser.add_option('-d', '--data', help=default('Dataset to use'), choices=['digits', 'text'], default='digits')
parser.add_option('-t', '--training', help=default('The size of the training set'), default=TRAIN_SET_SIZE, type="int")
parser.add_option('-f', '--features', help=default('Type of features'), default='basic', choices=['basic', 'pca', 'kmeans'])
parser.add_option('-i', '--iterations', help=default("Maximum iterations to run training"), default=50, type="int")
parser.add_option('-s', '--test', help=default("Amount of test data to use"), default=TEST_SET_SIZE, type="int")
parser.add_option('-n', '--neighbors', help=default("Number of clusters(for KMeans) or nearest neighbors(for KNN)"), default=5, type="int")
parser.add_option('-k', '--kernel', help=default("Kernel type"), default='linear', choices=['linear', 'rbf'])
parser.add_option('-C', '--C', help=default('Panelty C in SVM'), default=1.0, type='float')
parser.add_option('-m', '--dimension', help=default('Dimension reduction'), default=32, type='int')
parser.add_option('-v', '--visualize', help=default('Whether to visualize some results'), action='store_true')
options, otherjunk = parser.parse_args(argv)
if len(otherjunk) != 0: raise Exception('Command line input not understood: ' + str(otherjunk))
args = {}
# Set up variables according to the command line input.
print "Doing classification"
print "--------------------"
print "data:\t\t" + options.data
if options.data == 'digits':
print "classifier:\t\t" + options.classifier
else:
print "classifier:\t\t" + options.data + " classifier"
print "using enhanced features:\t" + str(options.features)
print "training set size:\t" + str(options.training)
if(options.data=="digits"):
if options.features == 'basic':
featureExtractor = BasicFeatureExtractorDigit()
elif options.features == 'pca':
featureExtractor = PCAFeatureExtractorDigit(options.dimension)
elif options.features == 'kmeans':
featureExtractor = KMeansClusterDigit(options.neighbors, options.iterations)
else:
featureExtractor = BasicFeatureExtractorDigit()
elif(options.data=="text"):
from sentimentAnalysis import FeatureExtractorText, ClassifierText
featureExtractor = FeatureExtractorText()
else:
print "Unknown dataset", options.data
print USAGE_STRING
sys.exit(2)
if options.data=="digits":
legalLabels = range(10)
else:
legalLabels = range(5)
if options.training <= 0:
print "Training set size should be a positive integer (you provided: %d)" % options.training
print USAGE_STRING
sys.exit(2)
if options.classifier == "perceptron":
classifier = classifiers.PerceptronClassifier(legalLabels, options.iterations)
elif options.classifier == "svm":
classifier = classifiers.SVMClassifier(legalLabels, options.iterations*100, options.C, options.kernel)
elif options.classifier == "knn":
classifier = classifiers.KNNClassifier(legalLabels, options.neighbors)
elif options.classifier == 'lr':
classifier = classifiers.LinearRegressionClassifier(legalLabels)
elif options.data == 'text':
classifier = ClassifierText(legalLabels)
else:
print "Do not use any classifier:", options.classifier
classifier = None
args['classifier'] = classifier
args['featureExtractor'] = featureExtractor
return args, options