def geterror(predictions, ytest):
# I have changed this geterror by the permission of the TA
return l2err(predictions, ytest)**2 / (2 * ytest.shape[0])
if __name__ == '__main__':
trainsize = 1000
testsize = 5000
#number of runs should be larger than 1 to make it possible to calculate standard deviation for standard error
numruns = 2
regressionalgs = {#'Random': algs.Regressor(),
# 'Mean': algs.MeanPredictor(),
# 'FSLinearRegression5': algs.FSLinearRegression({'features': [1,2,3,4,5]}),
'FSLinearRegression50': algs.FSLinearRegression({'features': range(385)}),
'RidgeLinearRegression': algs.RidgeLinearRegression({'features': range(385)}),
'LassoRegression' : algs.LassoRegression({'features': range(385)}),
'StochasticGradientDescent' : algs.StochasticGradientDescent({'features': range(385)}),
'BatchGradientDescent' : algs.BatchGradientDescent({'features': range(385)}),
'RMSProp' : algs.RMSProp({'features': range(385)}),
'AMSGrad' : algs.AMSGrad({'features': range(385)}),
}
numalgs = len(regressionalgs)
# Enable the best parameter to be selected, to enable comparison
# between algorithms with their best parameter settings
parameters = (
{
'regwgt': 0.01
},
# Can change this to other error values
return l2err(predictions, ytest) / ytest.shape[0]
if __name__ == '__main__':
trainsize = 1000
testsize = 5000
numruns = 1
regressionalgs = {
'Random':
algs.Regressor(),
'Mean':
algs.MeanPredictor(),
'FSLinearRegression5':
algs.FSLinearRegression({'features': [1, 2, 3, 4, 5]}),
'FSLinearRegression50':
algs.FSLinearRegression({'features': range(50)}),
'RidgeLinearRegression':
algs.RidgeLinearRegression(),
'Lasso':
algs.Lasso(),
'GSD':
algs.GSD(),
'BSD':
algs.BSD(),
}
numalgs = len(regressionalgs)
# Enable the best parameter to be selected, to enable comparison
# between algorithms with their best parameter settings
if __name__ == '__main__':
trainsize = 1000
testsize = 5000
numruns = 1
"""
By changing the value in feature variable it will change the number of features to be selected
"""
L = []
feature = 5
L.extend(range(feature))
regressionalgs = {#'Random': algs.Regressor(),
'Mean': algs.MeanPredictor(),
'FSLinearRegression5': algs.FSLinearRegression({'features': L}),
'FSLinearRegression50': algs.FSLinearRegression({'features': range(50)}),
'RidgeLinearRegression': algs.RidgeLinearRegression(),
'gradientDescent' : algs.gradientDescent(),
'stochasticgradientDescent':algs.stochasticgradientDescent(),
'LassoLinearRegression' : algs.LassoLinearRegression(),
'RMSPropRegression' : algs.RMSPropRegression(),
'amsGrad':algs.amsGrad(),
}
numalgs = len(regressionalgs)
# Enable the best parameter to be selected, to enable comparison
# between algorithms with their best parameter settings
parameters = (
#{'regwgt': 0.0},
{