面向增量式新类别发现的自适应发现与合并方法
Adaptive Discovering and Merging for Incremental Novel Class Discovery
AI总结：
针对终身学习中增量发现新类别易发生灾难性遗忘的问题，提出ADM范式，通过解耦发现过程、引入TC与PR约束实现自适应新类别发现，再经AMM混合结构无损失合并新知识，在class-iNCD任务上性能优于现有方法，还可助力class-IL任务缓解遗忘。
摘要：
终身学习的一项重要需求是能够从未标记数据中持续发现新类别。其核心挑战包含两个层面：既要发现并学习新类别，同时还要缓解对已有知识的灾难性遗忘问题。为此，我们提出了一种名为自适应发现与合并（Adaptive Discovering and Merging，ADM）的新范式，可在增量阶段自适应发现新类别，并在不影响原有知识的前提下将新知识整合到模型中。\n为实现自适应发现新类别，我们将表示学习与新类别发现解耦，采用三重比较（Triple Comparison，TC）和概率正则化（Probability Regularization，PR）对概率差异与多样性进行约束，以实现自适应的类别分配。为自适应合并已学习的新知识，我们提出了一种包含基部分支和新分支的混合结构，即自适应模型合并（Adaptive Model Merging，AMM）：它能降低新分支对旧类别的干扰以保留先前知识，还可在无性能损失、无参数增长的情况下将新分支合并到基础模型中。\n在多个数据集上开展的大量实验表明，ADM的性能显著优于现有的类别增量式新类别发现（class-incremental Novel Class Discovery，class-iNCD）方法。此外，我们提出的AMM还能通过缓解灾难性遗忘问题，为类别增量学习（class-incremental Learning，class-IL）任务带来增益。
英文摘要：
One important desideratum of lifelong learning aims to discover novel classes from unlabelled data in a continuous manner. The central challenge is twofold: discovering and learning novel classes while mitigating the issue of catastrophic forgetting of established knowledge. To this end, we introduce a new paradigm called Adaptive Discovering and Merging (ADM) to discover novel categories adaptively in the incremental stage and integrate novel knowledge into the model without affecting the original knowledge. To discover novel classes adaptively, we decouple representation learning and novel class discovery, and use Triple Comparison (TC) and Probability Regularization (PR) to constrain the probability discrepancy and diversity for adaptive category assignment. To merge the learned novel knowledge adaptively, we propose a hybrid structure with base and novel branches named Adaptive Model Merging (AMM), which reduces the interference of the novel branch on the old classes to preserve the previous knowledge, and merges the novel branch to the base model without performance loss and parameter growth. Extensive experiments on several datasets show that ADM significantly outperforms existing class-incremental Novel Class Discovery (class-iNCD) approaches. Moreover, our AMM also benefits the class-incremental Learning (class-IL) task by alleviating the catastrophic forgetting problem.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
