Question

是否可以使用单个函数替换下面的for循环，同时仍然生成字典输出？唯一的区别是X_train，y_train和X_test以及y_test。什么是最好的方式？

for _ in range(batch_count):
    rand_index = np.random.choice(len(X_train), size=config.batch_size)
    batch_X = X_train[rand_index].reshape((1, config.num_steps, config.input_size))
    batch_y = y_train[rand_index].reshape((1, config.num_steps, config.output_size)) 

    '''Each loop below completes one epoch training.'''

    train_data_feed = {inputs: batch_X, targets: batch_y, learning_rate: 0.0}



for _ in range(batch_count):
    rand_index = np.random.choice(len(X_train), size=config.batch_size)
    batch_X = X_test[rand_index].reshape((1, config.num_steps, config.input_size))
    batch_y = y_test[rand_index].reshape((1, config.num_steps, config.output_size)) 

    test_data_feed = {inputs: batch_X, targets: batch_y, learning_rate: 0.0}

Answer 1

你要求其他人为你做一个重构......这不是一个问题..

您可以创建一个名为reshape_data（x，y）的方法，它可以执行for循环：

def reshape_data(x, y, other parameters needed):
    for _ in range(batch_count):
        rand_index = np.random.choice(len(X_train), size=config.batch_size)
        batch_X = x[rand_index].reshape((1, config.num_steps, config.input_size))
        batch_y = y[rand_index].reshape((1, config.num_steps, config.output_size))

然后你可以从主代码中调用它：

reshape_data(X_train, y_train, batch_count, config)
reshape_data(X_test, y_test, batch_count, config)

当然，你可以进一步重构它。

PD：没有测试代码，它不起作用，因为变量的范围是config和batch_count等。

如何用单个函数替换两个for循环

1 个答案: