Efficient Memory Crystallization for Graph Learning under Non-Stationary Distribution Shifts

2026-10-02

AI Summary

This study addresses the high computational overhead and poor scalability of conventional generative memory methods for graph learning under non-stationary distribution shifts. We propose EMC, a training-free test-time adaptation framework that introduces a novel "crystallization" paradigm to eliminate generative memory. By leveraging closed-form distribution matching, EMC distills input domains into compact semantic memories, removing redundancy while modeling inter-domain dependencies. Coupled with a state evolution mechanism, this approach achieves tighter generalization error bounds without requiring any training. Experimental results demonstrate that EMC outperforms state-of-the-art baselines while reducing runtime and GPU memory consumption by 87.4% and 92.4%, respectively, thereby enabling efficient large-scale continual graph adaptation.

Abstract

Deep graph learning models deployed in real-world systems often need to cope with non-stationary environments, where the underlying graph distribution drifts continually over time. Prevailing solutions rely on training auxiliary generative modules to synthesize memory graphs for cross-domain adaptation, which incurs substantial computational overhead and scales poorly under prolonged distribution shifts. We argue that a more economical path exists: rather than generating memory, one can crystallize it. To this end, we propose Efficient Memory Crystallization (EMC), a training-free test-time framework that distills each incoming graph domain into a compact, semantically faithful memory through a closed-form solution to a memory-oriented distribution-matching objective, thereby eliminating redundant domain information under continual covariate shifts. To preserve both generalizability and adaptability as the model traverses a long sequence of target domains, EMC further models inter-domain dependencies through state-evolving memories and admits a theoretically grounded, tighter generalization error bound than direct adaptation. Extensive experiments demonstrate the superior performance of EMC over state-of-the-art baselines on graphs under non-stationary distribution shifts, while reducing average runtime by 87.4% and GPU memory consumption by 92.4% relative to the recent competitor, making continual graph adaptation practical at scale.

Problem

• Graph Learning
• Non-Stationary Distribution Shifts
• Continual Adaptation
• Memory Crystallization
• Computational Efficiency

Innovation

• Memory Crystallization
• Graph Learning
• Non-Stationary Distribution Shifts
• Training-free Adaptation
• Continual Domain Adaptation