研究人员开发了功能性任务网络（FTN），这是一种受哺乳动物新皮层启发的持续学习新方法。FTN使用自组织二元掩码来隔离不同任务的参数，防止灾难性遗忘，并在推理时实现无监督任务恢复。该方法在合成数据、带标签乱序的MNIST和置换MNIST上进行了测试，FTN-Slow遗忘率接近于零，而FTN-Fast则存在速度-保留权衡。另一篇论文探讨了依赖任务持续学习的理论恢复保证，分析了经验回放和知识蒸馏等范式。 AI
影响 FTN等持续学习方法的进步可以实现更强大、更适应性的AI系统，这些系统可以随着时间的推移进行学习而不会忘记过去的知识。
排序理由 该集群包含两篇arXiv论文，详细介绍了持续学习的新研究。
AI 生成摘要 · Google Gemini · 来自 5 个来源。 我们如何撰写摘要 →
arXiv:2604.27063v1 Announce Type: new Abstract: Continual learning agents with finite capacity must balance acquiring new knowledge with retaining the old. This requires controlled forgetting of knowledge that is no longer needed, freeing up capacity to learn. Weight decay, viewe…
arXiv:2604.24637v1 Announce Type: new Abstract: Block-sequential continual learning demands that a single model both protect prior solutions from catastrophic forgetting and efficiently infer at inference time which prior solution matches the current input without task labels. We…
Block-sequential continual learning demands that a single model both protect prior solutions from catastrophic forgetting and efficiently infer at inference time which prior solution matches the current input without task labels. We present Functional Task Networks (FTN), a param…
arXiv:2604.17578v2 Announce Type: replace Abstract: Continual learning (CL) is concerned with learning multiple tasks sequentially without forgetting previously learned tasks. Despite substantial empirical advances over recent years, the theoretical development of CL remains in i…
arXiv:2411.16073v3 Announce Type: replace-cross Abstract: Inspired by the \emph{Well-initialized Lottery Ticket Hypothesis (WLTH)}, we introduce Soft-Transformer (Soft-TF), a parameter-efficient framework for continual learning that leverages soft, real-valued subnetworks over a …