类别增量学习中的任务混淆与灾难性遗忘：面向判别式建模与生成式建模的数学框架
Task Confusion and Catastrophic Forgetting in Class-Incremental Learning: A Mathematical Framework for Discriminative and Generative Modelings
AI总结：
该研究构建类别增量学习数学框架，从理论上证明判别式建模因任务混淆无法达到最优，而生成式建模可克服该问题，并据此评估多种主流策略。
摘要：
在类别增量学习（class-IL）中，模型必须在测试时没有任务ID的情况下对所有先前见过的类别进行分类，从而导致任务混淆。尽管任务混淆是一项关键挑战，但其仍缺乏理论理解。我们提出了一个新颖的 class-IL 数学框架，并证明了不可行性定理，表明由于任务混淆，采用判别式建模不可能实现最优 class-IL。然而，我们建立了可行性定理，证明生成式建模能够通过克服任务混淆实现最优 class-IL。随后，我们利用该框架评估了流行的 class-IL 策略，包括正则化、偏差校正、回放和生成式分类器。我们的分析表明，采用生成式建模（无论是用于生成式回放还是直接分类（生成式分类器））对于最优 class-IL 都至关重要。
英文摘要：
In class-incremental learning (class-IL), models must classify all previously seen classes at test time without task-IDs, leading to task confusion. Despite being a key challenge, task confusion lacks a theoretical understanding. We present a novel mathematical framework for class-IL and prove the Infeasibility Theorem, showing optimal class-IL is impossible with discriminative modeling due to task confusion. However, we establish the Feasibility Theorem, demonstrating that generative modeling can achieve optimal class-IL by overcoming task confusion. We then assess popular class-IL strategies, including regularization, bias-correction, replay, and generative classifier, using our framework. Our analysis suggests that adopting generative modeling, either for generative replay or direct classification (generative classifier), is essential for optimal class-IL.
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。
