CRoNoS & MDA 2019: Start Registration
View Submission - CRONOSMDA2019
A0163
Title: Combining model and parameter uncertainty in Bayesian neural networks Authors:  Aliaksandr Hubin - NMBU (Norway) [presenting]
Abstract: Bayesian Neural Networks (BNNs) have recently regained a significant amount of attention in the deep learning community due to the development of scalable approximate Bayesian inference techniques for training them. The advantage of using BNNs is straightforward: parameter and prediction uncertainty become easily available allowing to perform rigid statistical analysis. However, so far there have been no scalable techniques capable of combining both model and parameter uncertainty developed. We introduce the concept of model uncertainty in Bayesian neural networks and hence make inference in the joint space of models and parameters. Furthermore, we suggest adaptation of a scalable variational inference approach with reparametrizations of marginal inclusion probabilities to incorporate the model space constraints.