Tensorflow - 总变异损失 - reduce_sum vs reduce_mean？

Question

为什么Total Variation Loss in Tensorflow建议使用reduce_sum而不是reduce_mean作为损失函数？

This can be used as a loss-function during optimization so as to suppress noise in images. If you have a batch of images, then you should calculate the scalar loss-value as the sum:
loss = tf.reduce_sum(tf.image.total_variation(images))

Answer 1

我联系了作者，似乎根本没有什么重要的原因。他提到也许 reduce_sum 对于他的测试用例比 reduce_mean 更有效，但鼓励我测试这两种情况并选择能给我最好结果的那个。