utils.py 文件源码

def data_shuffle(data_sets_org, percent_of_train, min_test_data=80, shuffle_data=False):
    """Divided the data to train and test and shuffle it"""
    perc = lambda i, t: np.rint((i * t) / 100).astype(np.int32)
    C = type('type_C', (object,), {})
    data_sets = C()
    stop_train_index = perc(percent_of_train[0], data_sets_org.data.shape[0])
    start_test_index = stop_train_index
    if percent_of_train > min_test_data:
        start_test_index = perc(min_test_data, data_sets_org.data.shape[0])
    data_sets.train = C()
    data_sets.test = C()
    if shuffle_data:
        shuffled_data, shuffled_labels = shuffle_in_unison_inplace(data_sets_org.data, data_sets_org.labels)
    else:
        shuffled_data, shuffled_labels = data_sets_org.data, data_sets_org.labels
    data_sets.train.data = shuffled_data[:stop_train_index, :]
    data_sets.train.labels = shuffled_labels[:stop_train_index, :]
    data_sets.test.data = shuffled_data[start_test_index:, :]
    data_sets.test.labels = shuffled_labels[start_test_index:, :]
    return data_sets