Open Saadmaghani opened 2 years ago
I believe that this library may only handle epistemic uncertainty currently. From section 2.1.2 of the TyXe paper:
At this time, we restrict the probabilistic model definition to weight space priors.
Weight uncertainty only accounts for the epistemic uncertainty of the model. To model the aleatoric uncertainty, one would need to incorporate a model for noise on either the inputs/outputs (or both!) a la Decomposition of Uncertainty in Bayesian Deep Learning for Efficient and Risk-sensitive Learning.
Firstly, thanks for the amazing repo. I'm a complete beginner and this repo really broke BNNs down simply in a very hands-on manner.
Could you guide me on how to calculate the uncertainty and break it into its epistemic & aleatoric components using TyXe?
Thanks!