thuml/predrnn-pytorch

Why is MSE calculated with "sum" instead of "mean"?

ANNMS-Projects opened this issue · 1 comments

Why is MSE calculated with "sum" instead of "mean"?
image

Is the amount of parameters of CrevNet really so small? It seems to be just the amount of parameters contained in a convolution of size 256x256x3x3 stored in float32.
image

Hi, we adopt the 'sum' following convention of previous paper. mean is also reasonable.

But, we don't know about you mentioned paper CrevNet. You can reach to their authors for this question.