semi-supervised-pytorch
semi-supervised-pytorch copied to clipboard
GaussianSample and GaussianMerge bug: positive log_var
As it's now it leads to log_var being positive only. But for standard case of standard normal distribution almost all q distribution log_var
s are less than zero. Or equals zero in case of N(0,1).
See here: https://github.com/wohlert/semi-supervised-pytorch/issues/16