Improving and Understanding Variational Continual Learning
This work addresses continual learning challenges for AI systems that need to learn tasks sequentially without forgetting, though it appears incremental as it builds on an existing competitive approach.
The authors tackled the problem of catastrophic forgetting and efficient model capacity usage in continual learning by improving the Variational Continual Learning (VCL) framework, achieving significantly better results on split MNIST and permuted MNIST benchmarks through a new best practice for mean-field variational Bayesian neural networks.
In the continual learning setting, tasks are encountered sequentially. The goal is to learn whilst i) avoiding catastrophic forgetting, ii) efficiently using model capacity, and iii) employing forward and backward transfer learning. In this paper, we explore how the Variational Continual Learning (VCL) framework achieves these desiderata on two benchmarks in continual learning: split MNIST and permuted MNIST. We first report significantly improved results on what was already a competitive approach. The improvements are achieved by establishing a new best practice approach to mean-field variational Bayesian neural networks. We then look at the solutions in detail. This allows us to obtain an understanding of why VCL performs as it does, and we compare the solution to what an `ideal' continual learning solution might be.