Dropout Inference in Bayesian Neural Networks with Alphadivergences
[edit]
Proceedings of the 34th International Conference on Machine Learning, PMLR 70:20522061, 2017.
Abstract
To obtain uncertainty estimates with realworld Bayesian deep learning models, practical inference approximations are needed. Dropout variational inference (VI) for example has been used for machine vision and medical applications, but VI can severely underestimates model uncertainty. Alphadivergences are alternative divergences to VI’s KL objective, which are able to avoid VI’s uncertainty underestimation. But these are hard to use in practice: existing techniques can only use Gaussian approximating distributions, and require existing models to be changed radically, thus are of limited use for practitioners. We propose a reparametrisation of the alphadivergence objectives, deriving a simple inference technique which, together with dropout, can be easily implemented with existing models by simply changing the loss of the model. We demonstrate improved uncertainty estimates and accuracy compared to VI in dropout networks. We study our model’s epistemic uncertainty far away from the data using adversarial images, showing that these can be distinguished from nonadversarial images by examining our model’s uncertainty.
Related Material


