Bayesian NN in torch

Is there a torch implementation of distribution layers, so it is possible to predict uncertainty in Neural Networks? Something similar to Tensorflow Probability or Yarin Gal's concrete dropout implementation.

Hi @fg.cordoni ,

We have support for a few distributions, and they can be use within nn_modules, here's an example:

This topic was automatically closed 21 days after the last reply. New replies are no longer allowed.

If you have a query related to it or one of the replies, start a new topic and refer back with a link.