标题： 基于无数据重放的连续人脸伪造检测中KAN局部性和特征漂移补偿投影的统一 显示英文标题

标题： Unifying Locality of KANs and Feature Drift Compensation Projection for Data-free Replay based Continual Face Forgery Detection

摘要： 快速发展的面部伪造技术要求检测器不断适应新的伪造方法，因此将面部伪造检测置于持续学习的范式中。 然而，当检测器学习新的伪造类型时，其在先前类型上的性能往往会迅速下降，这种现象称为灾难性遗忘。 Kolmogorov-Arnold 网络（KANs）使用局部可塑的样条作为其激活函数，使其能够通过仅修改函数的局部区域来学习新任务，同时不影响其他区域。 因此，它们天然适合解决灾难性遗忘问题。 然而，KANs 有两个显著的局限性：1）样条对于建模高维图像无效，而适用于图像的替代激活函数缺乏局部性的基本特性；2）在持续学习中，当不同领域的特征重叠时，由于对相同区域的重复修改，不同领域的映射到不同曲线区域总是会崩溃。 在本文中，我们提出了一种基于 KAN 的持续面部伪造检测（KAN-CFD）框架，包括一个领域分组 KAN 检测器（DG-KD）和一种无需使用先前任务数据的数据自由重放特征分离策略，通过 KAN 漂移补偿投影（FS-KDCP）。 DG-KD 使 KAN 能够适应高维图像输入，同时保留局部性和局部可塑性。 FS-KDCP 在不使用先前任务数据的情况下避免了 KAN 输入空间的重叠。 实验结果表明，所提出的方法在显著减少遗忘的同时实现了优越的性能。 显示英文摘要

摘要： The rapid advancements in face forgery techniques necessitate that detectors continuously adapt to new forgery methods, thus situating face forgery detection within a continual learning paradigm. However, when detectors learn new forgery types, their performance on previous types often degrades rapidly, a phenomenon known as catastrophic forgetting. Kolmogorov-Arnold Networks (KANs) utilize locally plastic splines as their activation functions, enabling them to learn new tasks by modifying only local regions of the functions while leaving other areas unaffected. Therefore, they are naturally suitable for addressing catastrophic forgetting. However, KANs have two significant limitations: 1) the splines are ineffective for modeling high-dimensional images, while alternative activation functions that are suitable for images lack the essential property of locality; 2) in continual learning, when features from different domains overlap, the mapping of different domains to distinct curve regions always collapses due to repeated modifications of the same regions. In this paper, we propose a KAN-based Continual Face Forgery Detection (KAN-CFD) framework, which includes a Domain-Group KAN Detector (DG-KD) and a data-free replay Feature Separation strategy via KAN Drift Compensation Projection (FS-KDCP). DG-KD enables KANs to fit high-dimensional image inputs while preserving locality and local plasticity. FS-KDCP avoids the overlap of the KAN input spaces without using data from prior tasks. Experimental results demonstrate that the proposed method achieves superior performance while notably reducing forgetting.