def mlhw6p17():
trainD = readTrainData()
testD = readTestData()
# training
forest = randomForest(trainD)
# testing
result = testEachTreeInForest(forest,testD)
eouts = map(lambda r:r["err"],result)
eouts.sort()
print "histogram"
his = histogram(eouts)
# Show histogram
for key in his:
print "[%3d%%,%3d%%) |%s %d (%.0f%%)" % (key,key+1,"|"*his[key],his[key],float(his[key])*100/len(eouts))
for i in xrange(len(forest)):
ein = testForest(forest[0:i+1],trainD)
eout = testForest(forest[0:i+1],testD)
ein = ein["err"]
eout = eout["err"]
print "%d,%.4f,%.4f" % (i+1,ein,eout)
def mlhw6p16():
trainD = readTrainData()
testD = readTestData()
dt = CART(trainD)
travel(dt)
eintest = testing(dt, trainD)
print eintest
eouttest = testing(dt, testD)
print eouttest
def mlhw6p16():
trainD = readTrainData()
testD = readTestData()
dt = CART(trainD)
travel(dt)
eintest = testing(dt,trainD)
print eintest
eouttest = testing(dt,testD)
print eouttest
def mlhw6p10():
trainD = readTrainData()
testD = readTestData()
G, U = adaboost(trainD, 300)
print "testing..."
# P10/11
for i in range(1, 300):
ein = adaboostTest(trainD, G[0:i])
eout = adaboostTest(testD, G[0:i])
print "%03d,%.6f,%.6f" % (i, ein, eout)
def mlhw6p10():
trainD = readTrainData()
testD = readTestData()
G,U = adaboost(trainD,300)
print "testing..."
# P10/11
for i in range(1,300):
ein = adaboostTest(trainD,G[0:i])
eout = adaboostTest(testD,G[0:i])
print "%03d,%.6f,%.6f" % (i,ein,eout)
def main():
CNNModel = classModel.Model()
directory = classModel.Directories()
files = glob.glob(directory.xmlPath)
para = classModel.Parameters()
modelPara = classModel.ModelParameters(len(files), para.lx, para.ly, 7)
y_train = readTrainLabels(directory, modelPara)
x_train = readTrainData(directory, modelPara, para)
CNNModel.model.fit(x_train,
y_train,
batch_size=modelPara.batch_size,
epochs=20,
validation_split=0.2,
callbacks=[directory.cp_callback])
def main():
directory = classModel.Directories()
files = glob.glob(directory.xmlPath)
para = classModel.Parameters()
modelPara = classModel.ModelParameters(len(files), para.lx, para.ly, 7)
CNNModel = classModel.Model()
CNNModel.model.compile(loss='categorical_crossentropy',
optimizer='adam',
metrics=['accuracy'])
CNNModel.model.load_weights(directory.checkpoint_path)
y_train = readTrainLabels(directory, modelPara)
x_train = readTrainData(directory, modelPara, para)
CNNModel.model.fit(x_train,
y_train,
batch_size=modelPara.batch_size,
epochs=20,
validation_split=0.2,
callbacks=[directory.cp_callback])
