explanation about the code you changed

kumar-shridhar / PyTorch-BayesianCNN

Bayesian Convolutional Neural Network with Variational Inference based on Bayes by Backprop in PyTorch.

MIT License

1.42k stars 323 forks source link

explanation about the code you changed #23

Closed ShellingFord221 closed 4 years ago

ShellingFord221 commented 5 years ago

Hi, could you please give a short explanation about the difference between old version and new version of the code? There are so many parameters in your old codes, especially in BBBlayers.py. I really don't know what does each parameter mean in old codes (e.g. qw_mean, conv_qw_mean, qw_logvar, conv_qw_std), and why should they be deleted in new version.

Thanks!

kumar-shridhar commented 5 years ago

Hi,

these parameters are what the name suggests: conv is for convolution and mean and variance means the mean and variance of the distributions for the layers. qw_mean and variance were somehow redundant and we removed them.