Abstract
We describe a limitation in the expressiveness of the predictive uncertainty
estimate given by mean-field variational inference (MFVI), a popular
approximate inference method for Bayesian neural networks. In particular, MFVI
fails to give calibrated uncertainty estimates in between separated regions of
observations. This can lead to catastrophically overconfident predictions when
testing on out-of-distribution data. Avoiding such overconfidence is critical
for active learning, Bayesian optimisation and out-of-distribution robustness.
We instead find that a classical technique, the linearised Laplace
approximation, can handle 'in-between' uncertainty much better for small
network architectures.
Users
Please
log in to take part in the discussion (add own reviews or comments).