Neural Network Ensembles and Variational Inference Revisited

Date December 2, 2018
Authors Marcin Tomczak, Siddharth Swaroop, Richard Turner

Ensembling methods and variational inference provide two orthogonal methods for obtain-ing reliable predictive uncertainty estimates for neural networks. In this work we compare and combine these approaches finding that: i) variational inference outperforms ensem-bles of neural networks, and ii) ensembled versions of variational inference bring further improvements. The first finding appears at odds with previous work (Lakshminarayananet al., 2017), but we show that the previous results were due to an ambiguous experimental protocol in which the model and inference method were simultaneously changed.

Share
,,

Related articles

Soft Q-Learning with Mutual-Information Regularization

研究分野
Papers

A Random Walk Approach to First-Order Stochastic Convex Optimization

研究分野
Papers

Bayesian Image Classification with Deep Convolutional Gaussian Processes

研究分野
Papers