wzm2256

I have spent the whole night trying to solve this problem..I am both glad and sad to know this is actually not my problem but just a bug...I will check...

@massimilianocomin Hi, I think it would be better to use 3D convolution than 1D with 3 channel, too much spatial information is lost in this 1D setting. I think what...

Let me explain why it is important. Because in generative modelling, we need to adjust the data and the Gaussian simultaneously to to match each other. Being biased, although just...

Please try debias=False and p=1.

:) You are welcome. This setting let you use entropic regularized (debias=False) EMD (p=1) loss, which is in accordant with the description in the library you mentioned.

The variance of the method is quite high. There is no way to avoid this. Try to repeat the experiment at least 10 times.

@GuHongyang Yes, but that is unfair. Because in practice you simply do not have any label to evaluate your results.

@huyong1369 I'm running related experiments using this codes. The average accuracy is about 81% for mnist. Try to repeat the experiment to achieve better results.

@Yijunmaverick Thanks, I've already figure it out. but thanks all the same.

Hi, it has been one year, has anyone solved this issue? > Does replacing create_window() with create_window(visible=False) solve your issue? In my case the program just becomes silent, no window,...