Marginal Replay vs Conditional Replay for Continual Learning
This addresses the challenge of error reduction in replay-based continual learning for classification tasks, though it appears incremental as it builds on existing replay paradigms.
The paper tackles the problem of continual classification learning by introducing 'conditional replay', a method that generates samples and labels together conditioned on class, reducing label inference errors. It demonstrates effectiveness on MNIST and FashionMNIST benchmarks, showing improvements over marginal replay and elastic weight consolidation.
We present a new replay-based method of continual classification learning that we term "conditional replay" which generates samples and labels together by sampling from a distribution conditioned on the class. We compare conditional replay to another replay-based continual learning paradigm (which we term "marginal replay") that generates samples independently of their class and assigns labels in a separate step. The main improvement in conditional replay is that labels for generated samples need not be inferred, which reduces the margin for error in complex continual classification learning tasks. We demonstrate the effectiveness of this approach using novel and standard benchmarks constructed from MNIST and FashionMNIST data, and compare to the regularization-based \textit{elastic weight consolidation} (EWC) method.