if __name__ == '__main__':
trainsize = 1000
testsize = 5000
numruns = 10
classalgs = { #'Random': algs.Classifier(),
# 'Naive Bayes': algs.NaiveBayes({'usecolumnones': False}),
# 'Naive Bayes Ones': algs.NaiveBayes({'usecolumnones': True}),
# 'Linear Regression': algs.LinearRegressionClass(),
# 'Logistic Regression': algs.LogitReg(),
# 'L1 Logistic Regression': algs.LogitReg({'regularizer': 'l1'}),
# 'L2 Logistic Regression': algs.LogitReg({'regularizer': 'l2'}),
# 'Elastic Logistic Regression': algs.LogitReg({'regularizer': 'elastic'}),
# 'Logistic Alternative': algs.LogitRegAlternative(),
'Neural Network': algs.NeuralNet({'epochs': 100})
}
numalgs = len(classalgs)
parameters = (
{
'regwgt': 0.0,
'nh': 4
},
{
'regwgt': 0.01,
'nh': 8
},
{
'regwgt': 0.05,
'nh': 16
'stepsize': 0.001
}),
'Logistic Regression':
algs.LogitReg({
'lamb': 0.001,
'stepsize': 0.001
}),
'kernel Logistic Regression':
algs.KernelLogitReg({'k': 30}),
'Hamming kernel Logistic Regression':
algs.KernelLogitReg({
'kernel': 'hamming',
'k': 20
}),
'Neural Network':
algs.NeuralNet({'epochs': 100}),
'Neural Network2':
algs.NeuralNet2({'epochs': 100})
}
numalgs = len(classalgs)
cls = {
'Logistic RegressionRegularized':
algs.LogitReg({
'regularizer': 'l2',
'lamb': 0.001,
'stepsize': 0.001
}),
'Logistic RegressionRegularized':
algs.LogitReg({
'regularizer': 'l2',
errors = {}
# accuracies = []
classalgs = {}
numparams = 0
parameters = {}
for i in range(5):
smallDataSet = dataset[i * 1300:(i + 1) * 1300]
classalgs = {
'Naive Bayes':
algs.NaiveBayes({'usecolumnones': False}),
'Logistic Regression':
algs.LogitReg(),
'Neural Network':
algs.NeuralNet({
'epochs': 100,
'stepsize': 0.01,
'nh': 8,
'ni': 19
})
# 'L1 Logistic Regression': algs.LogitReg({'regularizer': 'l1'}),
# 'L2 Logistic Regression': algs.LogitReg({'regularizer': 'l2'}),
}
numalgs = len(classalgs)
parameters = (
# {'regwgt': 0.0, 'nh': 4},
{
'regwgt': 0.01,
'nh': 8
},
# {'regwgt': 0.05, 'nh': 16},
# {'regwgt': 0.1, 'nh': 32},
)
if __name__ == '__main__':
trainsize = 5000
testsize = 5000
numruns = 5
classalgs = {
'Random':
algs.Classifier(),
#'Naive Bayes': algs.NaiveBayes({'usecolumnones': False}),
#'Naive Bayes Ones': algs.NaiveBayes({'usecolumnones': True}),
#'Linear Regression': algs.LinearRegressionClass({'regularizer': 'l2', 'regwgt': 0.0}),
#'Logistic Regression': algs.LogitReg({'regularizer': 'l2', 'regwgt': 0.0}),
'Neural Network':
algs.NeuralNet({
'epochs': 150,
'transfer': 'sigmoid',
'stepsize': 0.1,
'nh': 32
}),
#'Kernel Logistic Regression': algs.KernelLogitReg({'kernel': 'None', 'num_centers': 10, 'regularizer': 'None', 'regwgt': 0.0}),
#'Linear Kernel Logistic Regression': algs.KernelLogitReg({'kernel': 'linear', 'num_centers': 100, 'regularizer': 'l2', 'regwgt': 0.0}),
#'Hamming Kernel Logistic Regression': algs.KernelLogitReg({'kernel': 'hamming', 'num_centers': 100, 'regularizer': 'l2', 'regwgt': 0.0})
}
numalgs = len(classalgs)
parameters = (
#{'regwgt': 0.0, 'nh': 4},
#{'regwgt': 0.01, 'nh': 8},
#{'regwgt': 0.05, 'nh': 16},
#{'regwgt': 0.1, 'nh': 32},
{
'regwgt': 0.0,
testsetx = np.hstack((datasettest, np.ones((datasettest.shape[0], 1))))
testset = (testsetx, testlab)
return trainset, testset
if __name__ == '__main__':
trainset, testset = loadsusy()
#trainset, testset = loadmadelon()
print('Running on train={0} and test={1} samples').format(
trainset[0].shape[0], testset[0].shape[0])
nnparams = {'ni': trainset[0].shape[1], 'nh': 64, 'no': 1}
classalgs = {
'Random': algs.Classifier(),
'Linear Regression': algs.LinearRegressionClass(),
'Naive Bayes': algs.NaiveBayes({'usecolumnones': False}),
'Naive Bayes Ones': algs.NaiveBayes(),
'Logistic Regression': algs.LogitReg(),
'Neural Network': algs.NeuralNet(nnparams),
'Linear Class': algs.LinearClass()
}
for learnername, learner in classalgs.iteritems():
print 'Running learner = ' + learnername
# Train model
learner.learn(trainset[0], trainset[1])
# Test model
predictions = learner.predict(testset[0])
accuracy = getaccuracy(testset[1], predictions)
print 'Accuracy for ' + learnername + ': ' + str(accuracy)
else:
k_foldStratified = 0
k_foldClass = 0
classalgs = {
'Random':
algs.Classifier(),
# 'Naive Bayes': algs.NaiveBayes({'usecolumnones': False}),
# 'Naive Bayes Ones': algs.NaiveBayes({'usecolumnones': True}),
'Linear Regression':
algs.LinearRegressionClass(),
'Logistic Regression':
algs.LogitReg(),
'Neural Network':
algs.NeuralNet({'hiddenLayers': 1}),
'Neural Network':
algs.NeuralNet({
'epochs': 100,
'hiddenLayers': 2
}),
'Kernel Logistic Regression linear':
algs.KernelLogitReg({
'kernel': 'linear',
'regularizer': 'l2'
}),
'Kernel Logistic Regression hamming':
algs.KernelLogitReg({
'kernel': 'hamming',
'regularizer': 'l2'
})
fold=10
trainlabel=np.reshape(trainset[1],(-1,1))
trset = np.hstack((trainset[0],trainlabel))
numinputs = trset.shape[1]-1
np.random.shuffle(trset)
parts = [trset[i::fold] for i in xrange(fold)]
obj=[]
print('Running on train={0} and test={1} samples').format(trainset[0].shape[0], testset[0].shape[0])
parm_pass={'Neural Network':{'ni': trset.shape[1]-1, 'nh': 0, 'no': 1},
'Logistic Regression':{'regwt':0,'type':"L2"}}
classalgs = {'Linear Regression': algs.LinearRegressionClass(),
'Naive Bayes Ones': algs.NaiveBayes(),
'Logistic Regression': algs.LogitReg(parm_pass['Logistic Regression']),
'Neural Network': algs.NeuralNet(parm_pass['Neural Network'])
}
classalgs1 = collections.OrderedDict(sorted(classalgs.items()))
best_parm=[]
for learnername , learner in classalgs1.iteritems():
print 'Running learner = ' + learnername
# # Train model
parm_accuracy={}
for j in range(0,len(parm_dict[learnername])):
parm=[]
