BNN - Loss vs Accuracy?

afaict that guide is incorrect so i wouldn’t suggest that you base your code on it.

i haven’t vetted it but this is probably a better place to look.

in general, bayesian neural networks are an active area of research and may not be a great entry point for non-experts. (also, they generally give poor empirical performance—at least unless you apply all kinds of expert tweaks and even then the performance is generally poor)