Tag
提出REDAR：将类增量学习中的正则化强度选择建模为序列决策问题，用轻量级RL智能体根据当前性能与数据统计动态输出每个阶段的知识蒸馏系数，在CIFAR-100和Tiny-ImageNet上降低了遗忘并提升了前向迁移。
该论文对九种主流类增量学习（CIL）方法在五个冻结的蛋白质编码器上的表现进行了系统基准测试，发现没有方法能可靠跨编码器迁移，并分析了失败源于嵌入几何异质性与不稳定的梯度稳定性信号，主张持续学习设计应将表征本身作为核心组件。
This paper proposes a lightweight plastic-memory framework for graph few-shot class-incremental learning, which uses an evolving micro-clustering structure and meta-learning to balance knowledge retention and adaptability to new classes with limited data.
This paper proposes FedFMX, a Fisher-Routed Mixture of Experts framework for Federated Class-Incremental Learning, addressing capacity conflict, catastrophic forgetting, and data heterogeneity via adaptive expert specialization.
This paper introduces CaRE, a novel continual learning framework using a bi-level routing mixture-of-experts mechanism to effectively handle class-incremental learning over sequences of 300+ tasks.