def ann_visual_weights(X_train, y_train, X_test, y_test):
xes, ycv, y_c = range(1,6), [], []
for m in xes:
layers = [ X_train.shape[1] ] + [ 30 ] * m + [ y_train.shape[1] ]
cost, cv_cost = test_ann(X_train, y_train, X_test, y_test, layers, 1000)
y_c.append(cost)
ycv.append(cv_cost)
pl.plot(xes, y_c, 'r-')
pl.plot(xes, ycv, 'b-')
pl.show()
return
# y_valid = np.array(y_valid)
print X_test.shape, y_test.shape, \
X_train.shape, y_train.shape #, \
# X_valid.shape, y_valid.shape
return X_train, y_train, X_test, y_test # , X_valid, y_valid
if __name__ == '__main__':
if sys.argv[1] == '-nt':
X_train, y_train, X_test, y_test = input_data("./data/", test_samples=[7,8])
test_ann(X_train, y_train, X_test, y_test, [X_train.shape[1], 30, y_train.shape[1]])
elif sys.argv[1] == '-nv':
X_train, y_train, X_test, y_test = input_data("./data/", test_samples=[7,8])
ann_visual_weights(X_train, y_train, X_test, y_test)
elif sys.argv[1] == '-rt':
X_train, y_train, X_test, y_test = input_data("./data/", test_samples=[])
test_rbm( X_train, y_train, X_test, y_test)
