Analytic Drift Resister for Non-Exemplar Continual Graph Learning

2026-04-02

AI Summary

This work addresses the feature drift caused by retaining only class prototypes and the limited plasticity of analytic continual learning models in exemplar-free continual graph learning. To overcome these limitations, the authors propose the Analytic Drift Resister (ADR) framework, which enhances model plasticity by iteratively applying backpropagation to circumvent the typical freezing of pretrained models. ADR further integrates hierarchical analytic consolidation with an analytically reconstructed classifier to fully suppress feature drift and achieve theoretically zero forgetting. As the first method to attain zero forgetting in exemplar-free continual graph learning, ADR matches or surpasses state-of-the-art performance on four node classification benchmarks, effectively balancing stability against forgetting with adaptability to new tasks.

Abstract

Non-Exemplar Continual Graph Learning (NECGL) seeks to eliminate the privacy risks intrinsic to rehearsal-based paradigms by retaining solely class-level prototype representations rather than raw graph examples for mitigating catastrophic forgetting. However, this design choice inevitably precipitates feature drift. As a nascent alternative, Analytic Continual Learning (ACL) capitalizes on the intrinsic generalization properties of frozen pre-trained models to bolster continual learning performance. Nonetheless, a key drawback resides in the pronounced attenuation of model plasticity. To surmount these challenges, we propose Analytic Drift Resister (ADR), a novel and theoretically grounded NECGL framework. ADR exploits iterative backpropagation to break free from the frozen pre-trained constraint, adapting to evolving task graph distributions and fortifying model plasticity. Since parameter updates trigger feature drift, we further propose Hierarchical Analytic Merging (HAM), performing layer-wise merging of linear transformations in Graph Neural Networks (GNNs) via ridge regression, thereby ensuring absolute resistance to feature drift. On this basis, Analytic Classifier Reconstruction (ACR) enables theoretically zero-forgetting class-incremental learning. Empirical evaluation on four node classification benchmarks demonstrates that ADR maintains strong competitiveness against existing state-of-the-art methods.

Problem

• Non-Exemplar Continual Graph Learning
• feature drift
• catastrophic forgetting
• model plasticity
• Graph Neural Networks

Innovation

• Analytic Drift Resister
• Non-Exemplar Continual Graph Learning
• Hierarchical Analytic Merging
• Feature Drift Resistance
• Zero-Forgetting Learning