1. Two types of padding in TensorFlow, SAME and VALID
In short: padding='SAME' means that padding 0 at both ends is used for completion, and the number of padding 0 on the left and right may be different.
padding='VALID' indicates that the method of discarding is adopted, and the redundant elements are directly discarded without padding.
For details, please refer to: http://blog.csdn.net/wuzqchom/article/details/74785643