Unifying Locality of KANs and Feature Drift Compensation for Data-free Continual Face Forgery Detection

2025-08-05

AI Summary

To address catastrophic forgetting in data-free continual learning for face forgery detection, this work identifies two key challenges inherent to Kolmogorov–Arnold Networks (KANs): insufficient expressivity of local spline activation functions for high-dimensional image modeling, and mapping region collapse caused by cross-domain feature overlap. To overcome these, we propose a domain-grouped KAN framework featuring: (i) plastic spline activation functions that jointly preserve locality and adapt to image semantics; and (ii) a history-data-free feature separation mechanism integrated with drift-compensated projection to mitigate input-space overlap. Extensive experiments demonstrate that our method significantly alleviates forgetting across multi-stage continual learning scenarios, consistently outperforming state-of-the-art approaches on mainstream benchmarks.

Technology Category

• Computer Vision: Generative Adversarial Networks (GANs) for Vision
• Machine Learning: Life-Long and Continual Learning
• Cognitive Modeling & Cognitive Systems: Conceptual Inference and Reasoning

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

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.

Problem

• Address catastrophic forgetting in continual face forgery detection
• Enable KANs to model high-dimensional images with locality
• Prevent feature drift in continual learning without prior data

Innovation

• Domain-Group KAN Detector for high-dimensional images
• Data-free Feature Separation via KAN Drift Compensation
• Locality-preserving KANs for continual learning