Information-theoretic Online Memory Selection for Continual Learning
This work addresses memory selection for continual learning systems, offering incremental improvements in handling data streams.
The paper tackles the problem of online memory selection in task-free continual learning by proposing surprise and learnability criteria from an information-theoretic perspective, resulting in improved robustness against data imbalance and demonstrated efficiency on benchmarks.
A challenging problem in task-free continual learning is the online selection of a representative replay memory from data streams. In this work, we investigate the online memory selection problem from an information-theoretic perspective. To gather the most information, we propose the \textit{surprise} and the \textit{learnability} criteria to pick informative points and to avoid outliers. We present a Bayesian model to compute the criteria efficiently by exploiting rank-one matrix structures. We demonstrate that these criteria encourage selecting informative points in a greedy algorithm for online memory selection. Furthermore, by identifying the importance of \textit{the timing to update the memory}, we introduce a stochastic information-theoretic reservoir sampler (InfoRS), which conducts sampling among selective points with high information. Compared to reservoir sampling, InfoRS demonstrates improved robustness against data imbalance. Finally, empirical performances over continual learning benchmarks manifest its efficiency and efficacy.