technique

Class-Incremental Learning

techniqueactiveprovisionalclass-incremental-learning-64f9769b·1 events·first seen 20d ago

Aliases: Class-Incremental Learning

Co-occurring entities

Optimal Transport AREA Principal Geodesic Analysis Variational Information Bottleneck ICML 2026 LAMDA-CL CLIP

More like this (12)

Continual Learning Reinforcement Learning for Code Operator Learning Multiple Instance Learning (MIL)In-Context Multiple Instance Learning Iterated Amplification Multimodal Continual Instruction Tuning Personalized Evaluation as Learning Structured Interactive Learning curriculum learning Alternating Token-Weighted Unlearning Q-learning

Recent events (1)

5arXiv · cs.LG·20d ago·source ↗

AREA: Attribute Extraction and Aggregation for CLIP-Based Class-Incremental Learning

AREA is a new method for CLIP-based Class-Incremental Learning (CIL) that decomposes the classification process into attribute extraction and aggregation stages to combat catastrophic forgetting. Extraction is stabilized by anchoring visual and textual attributes on a hyperspherical embedding space via principal geodesic analysis, while aggregation uses lightweight task-specific experts regularized by a variational information bottleneck. Inference employs optimal transport routing over task attribute manifolds. The method is reported to consistently outperform state-of-the-art CIL approaches and is accepted at ICML 2026.

Evaluation and Benchmarking Multimodal Progress Optimal Transport AREA Principal Geodesic Analysis +5 more