Hierarchical Indian buffet neural networks for Bayesian continual learning

Conference item

We place an Indian Buffet process (IBP) prior over the structure of a Bayesian Neural Network (BNN), thus allowing the complexity of the BNN to increase and decrease automatically. We further extend this model such that the prior on the structure of each hidden layer is shared globally across all layers, using a Hierarchical-IBP (H-IBP). We apply this model to the problem of resource allocation in Continual Learning (CL) where new tasks occur and the network requires extra resources. Our model uses online variational inference with reparameterisation of the Bernoulli and Beta distributions, which constitute the IBP and H-IBP priors. As we automatically learn the number of weights in each layer of the BNN, overfitting and underfitting problems are largely overcome. We show empirically that our approach offers a competitive edge over existing methods in CL.

Authors: Kessler, S; Nguyen, V; Zohren, S; Roberts, SJ

• Publication status: Published
• Peer review status: Peer reviewed
• Publisher: PMLR
• Host title: Proceedings of the Thirty-Seventh Conference on Uncertainty in Artificial Intelligence
• Pages: 749-759
• Publication date: 2021-12-01
• Acceptance date: 2021-05-12
• Event title: 37th Conference on Uncertainty in Artificial Intelligence (UAI 2021)
• Event location: Virtual event
• Event website: https://www.auai.org/uai2021/
• Event start date: 2021-07-27
• Event end date: 2021-07-30
• EISSN: 2640-3498
• ISSN: 2640-3498
• Language: English