Variational continual learning
Accepted version
Peer-reviewed
Repository URI
Repository DOI
Change log
Authors
Turner, RE
Bui, Thang
Li, Yingzhen
Cuong, Nguyen
Abstract
This paper develops variational continual learning (VCL), a simple but general framework for continual learning that fuses online variational inference (VI) and recent advances in Monte Carlo VI for neural networks. The framework can suc- cessfully train both deep discriminative models and deep generative models in complex continual learning settings where existing tasks evolve over time and en- tirely new tasks emerge. Experimental results show that VCL outperforms state- of-the-art continual learning methods on a variety of tasks, avoiding catastrophic forgetting in a fully automatic way.
Description
Keywords
Journal Title
Conference Name
International Conference on Learning Representations
Journal ISSN
Volume Title
Publisher
Publisher DOI
Sponsorship
Engineering and Physical Sciences Research Council (EP/L000776/1)
Engineering and Physical Sciences Research Council (EP/M026957/1)
Engineering and Physical Sciences Research Council (EP/M026957/1)
EPSRC grants EP/M0269571 and EP/L000776/1