Coresets via Bilevel Optimization for Continual Learning and Streaming
This addresses the problem of efficiently handling large data streams under resource constraints for machine learning practitioners, representing a novel method for a known bottleneck.
The paper tackles the limitation of existing coreset constructions to simple models by proposing a novel coreset construction via cardinality-constrained bilevel optimization, enabling efficient coresets for deep neural networks and demonstrating empirical benefits in continual learning and streaming settings.
Coresets are small data summaries that are sufficient for model training. They can be maintained online, enabling efficient handling of large data streams under resource constraints. However, existing constructions are limited to simple models such as k-means and logistic regression. In this work, we propose a novel coreset construction via cardinality-constrained bilevel optimization. We show how our framework can efficiently generate coresets for deep neural networks, and demonstrate its empirical benefits in continual learning and in streaming settings.