🤖 AI Summary
In continual learning, catastrophic forgetting severely degrades model performance under extremely small replay buffers (1–3 samples per class), and existing sample selection methods exhibit limited effectiveness under such stringent memory constraints. To address this, we propose Typicality-Driven Experience Replay (TEAL), the first method that jointly leverages unsupervised feature-space density estimation and intra-class typicality scoring as the core criterion for sample selection. TEAL requires no additional training, introduces no architectural modifications, and is plug-and-play compatible with mainstream replay frameworks—including ER, DER, and SCR. Evaluated on multiple class-incremental learning benchmarks under minimal memory budgets, TEAL achieves statistically significant gains in average accuracy over prior selection strategies, attaining state-of-the-art performance. It substantially enhances historical knowledge retention within highly constrained buffer capacities.
📝 Abstract
Continual Learning is an unresolved challenge, whose relevance increases when considering modern applications. Unlike the human brain, trained deep neural networks suffer from a phenomenon called catastrophic forgetting, wherein they progressively lose previously acquired knowledge upon learning new tasks. To mitigate this problem, numerous methods have been developed, many relying on the replay of past exemplars during new task training. However, as the memory allocated for replay decreases, the effectiveness of these approaches diminishes. On the other hand, maintaining a large memory for the purpose of replay is inefficient and often impractical. Here we introduce TEAL, a novel approach to populate the memory with exemplars, that can be integrated with various experience-replay methods and significantly enhance their performance with small memory buffers. We show that TEAL enhances the average accuracy of existing class-incremental methods and outperforms other selection strategies, achieving state-of-the-art performance even with small memory buffers of 1-3 exemplars per class in the final task. This confirms our initial hypothesis that when memory is scarce, it is best to prioritize the most typical data. Code is available at this https URL: https://github.com/shahariel/TEAL.