def trial_running_pca():
lower_dimension = 4
iris_dataset = IRIS(crossvalidation=1)
test_data,test_label,train_data,train_label = iris_dataset.load_data()
test_points,test_dimension = test_data.shape
train_points,train_dimension = train_data.shape
complete_data = zeros((test_points+train_points,test_dimension),float32)
complete_data[:test_points,:] = test_data
complete_data[test_points:,:] = train_data
'''
A = LDA(complete_data.T)
dimension_reduced_data = dot(complete_data,A)
markers = {'1':'r.','2':'g.','3':'b.'}
plot_2d(dimension_reduced_data,test_label.tolist()+train_label.tolist(),markers)
'''
pca = PCA(train_data, output_dim=lower_dimension)
dimension_reduced_test_data = pca.execute(test_data,n=lower_dimension)
dimension_reduced_train_data = pca.execute(train_data,n=lower_dimension)
train_data = train_data.T
test_data = test_data.T
dimension_reduced_test_data = dimension_reduced_test_data.T
dimension_reduced_train_data = dimension_reduced_train_data.T
classification_error(dimension_reduced_train_data, train_label, dimension_reduced_test_data, test_label )
print train_data.shape
print test_data.shape
print dimension_reduced_test_data.shape
markers = {'1':'r.','2':'g.','3':'b.', '4':'b*', '5':'g*', '6':'r*','7':'rs', '8':'gs','9':'bs'}
plot_2d(dimension_reduced_test_data,test_label.tolist(),markers)
def run_kde():
print "-- Starting kde--"
landsat_dataset = IRIS(crossvalidation = 0)
test_data,test_label,train_data,train_label = landsat_dataset.load_data()
print test_data.shape
print train_data.shape
lower_dimension = 2
kde_cub = KDECUB(train_data, train_label, lower_dimension, test_data = test_data, test_label = test_label)
kde_cub.train()