def preprocess_data(X,y,nb_classes=10, binarize=False, noise_prop=0.49): X_new = X.reshape(-1, 784) X_new = X_new.astype('float32') print(X_new.shape[0], 'samples') if binarize: X_new = binaryze_dataset(X_new, threshold=0.5) if noise_prop != None: print("Adding salt and pepper ({})".format(noise_prop)) X_new = add_salt_and_pepper(X_new,proportion=noise_prop) if nb_classes == 2: Y = numpy.in1d(y,[0,2,4,6,8]).astype('float64') else: Y = y print(Y.shape) return X_new,Y
def preprocess_data(X,y,nb_classes=10, binarize=False, noise=False, proportion=0.1): X_new = X.reshape(-1, 784) X_new = X_new.astype('float32') X_new /= 255.0 print(X_new.shape[0], 'samples') if binarize: X_new = binaryze_dataset(X_new, threshold=0.5) if noise: X_new = add_salt_and_pepper(X_new,proportion=proportion) if nb_classes == 2: Y = np.in1d(y,[0,2,4,6,8]).astype('float64') else: Y = np_utils.to_categorical(y,nb_classes) return X_new,Y