Entangled No More: Multi-Domain Decoupling for Robust Dynamic Graph Neural Networks

Dynamic graphs are pervasive in real-world systems, but their tightly entangled spatiotemporal evolution causes significant modeling challenges. Existing Dynamic Graph Neural Networks (DGNNs) lack a principled framework for systematically decoupling this multi-domain entanglement, raising two key problems: (i) representation drift caused by structural incompleteness, and (ii) signal distortion amplified by noise perturbation. These problems can accumulate over time, forming temporal redundancy that weakens robustness of DGNNs. In view of these, we propose DeR-Mamba(Decoupling for Robust Mamba), a multi-domain decoupling framework for robust DGNNs. To address (i), we develop the Multi-Particle Kernel Kalman observation field (MP-K$^2$alman), which achieves spatial decoupling by sampling latent evolution paths in kernel subspaces and performing Kalman-style updates to estimate structural states. To address (ii), we design the Adversarial-aware Frequency Decoupling Module (AFDM), which performs frequency-domain decoupling and dynamic cross-frequency modulation to purify spectral signals. Finally, a self-consistent dynamic graph state-space system performs temporal decoupling to control redundancy, suppressing residual disturbances through discretized cross-time modeling and selective snapshot scanning. Extensive experiments on benchmark datasets with adversarial attacks validate its superior robustness.

Many real-world systems, such as social networks and financial markets, are highly interconnected and constantly changing. Artificial Intelligence uses "dynamic graphs" to track these complex systems, but current models often struggle because the changes happening across space and time are densely tangled together. This entanglement causes two major problems: the AI gets confused when parts of the network's structure are incomplete, and it gets easily misled by noisy or malicious data. Over time, these errors pile up, making predictions unreliable. To solve this, we developed DeR-Mamba, a new framework that acts as a systematic untangler. It uses a smart estimation technique to fill in the blanks when structural information is missing, acts as an advanced filter to separate useful information from corrupted data, and carefully tracks changes over time to prevent past errors from accumulating. Our tests show that DeR-Mamba makes AI models significantly tougher and more reliable, allowing them to perform accurately even when faced with heavily corrupted data or intentional attacks.