Time-varying Interaction Graph ODE for Dynamic Graph Representation Learning

Xiaoyi Wang, Zhiqiang Wang, Jianqing Liang, Xingwang Zhao, Chuangyin Dang, Zhen Jin, Jiye Liang · 2026

Read Time-varying Interaction Graph ODE for Dynamic Graph Representation Learning by Xiaoyi Wang, Zhiqiang Wang, Jianqing Liang, Xingwang Zhao, Chuangyin Dang, Zhen Jin, Jiye Liang (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-24811.txt.

Introduction

Wang and six colleagues propose TI-ODE for dynamic graph learning. Existing graph ODEs use one message function for all interactions. That misses the diversity of interaction patterns over time. TI-ODE splits evolution into learnable basis functions per interaction type. Time-dependent weights combine them as patterns evolve. On six dynamic datasets it beats existing methods on attribute prediction. Theory and tests also show stronger robustness than unified models.

Catalog record

CREATORXiaoyi Wang, Zhiqiang Wang, Jianqing Liang, Xingwang Zhao, Chuangyin Dang, Zhen Jin, Jiye Liang
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE78574
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2604.24811

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Time-varying Interaction Graph ODE for Dynamic Graph Representation Learning Xiaoyi Wang, Zhiqiang Wang, Jianqing Liang, Xingwang Zhao, Chuangyin Dang, Zhen Jin, Jiye Liang Abstract Graph neural Ordinary Differential Equations (ODE) combine neural ODE with the message passing mechanism of Graph Neural Networks (GNN), providing a continuous-time modeling method for graph representation learning. However, in dynamic graph scenarios, existing graph neural ODEs typically employ a unified message passing mechanism, assuming that inter-node interactions share the same message passing function at any time, which makes it challenging to capture the diversity and time-varying nature of inter-node interaction patterns. To address this, we propose Time-varying Interaction Graph Ordinary Differential Equations (TI-ODE). The core idea of TI-ODE is to decompose the evolution function of a graph ODE into a set of learnable interaction basis functions, where each basis function corresponds to a distinct type of inter-node interaction. These basis functions are dynamically combined through time-dependent learnable weights, enabling inter-node interaction patterns to adaptively evolve over time. Experimental results on six dynamic graph datasets demonstrate that TI-ODE consistently outperforms existing methods and achieves state-of-the-art performance on attribute prediction tasks, and experiments on the Covid dataset further verify the interpretability and generalizability of our TI-ODE. Furthermore, we demonstrate both theoretically and empirically that TI-ODE exhibits superior robustness compared to models utilizing a unified message-passing mechanism. keywords Graph Neural Networks ,Dynamic Graph ,Neural Ordinary Differential Equations ,Deep Learning † † credit: Conceptualization of this study, Formal analysis, Methodology, Resources, Software, Data curation, Validation, Writing - Original draft preparation † † credit: Conceptualization of this study, Funding acquisition, Methodology, Project administration, Supervision, Writing – review and editing † † credit: Conceptualization of this study, Funding acquisition, Methodology, Project administration, Supervision, Writing – review and editing † † credit: Methodology, Supervision, Writing – review and editing † † credit: Methodology, Supervision, Writing – review and editing † † credit: Methodology, Supervision, Writing – review and editing † † credit: Funding acquisition, Writing – review and editing, Project administration, Supervision † † corresponding: Corresponding author † † note: This manuscript is under review by Elsevier. 1 Introduction Dynamic graphs are ubiquitous in real-world systems such as social networks, transportation networks, and biological networks, where nodes and edges evolve over time and exhibit complex dynamic behaviors. To model such network dynamics, Dynamic Graph Neural Networks (DGNN) ( 42 ; 44 ) have been proposed. Building upon Graph Neural Networks (GNN) ( 29 ) , DGNN explicitly incorporate the temporal dimension to characterize the dynamic evolution of both node states and graph structures. However, traditional DGNN based on discrete neural architectures ( 44 ) rely on predefined time steps and fixed hierarchical update mechanisms, which limits their ability to accurately capture continuous-time dynamics. Furthermore, such fixed data-processing pipelines reduce model adaptability, making it difficult to handle irregularly observed data. The redundant structures and parameters inherent in discrete-time designs also lead to increased computational and memory overhead ( 13 ) . To address these limitations, neural Ordinary Differential Equations (ODE) ( 3 ) offer a continuous-time modeling paradigm that characterizes the system’s continuous evolution by parameterizing the derivative of states with respect to time. Researchers have further extended this continuous modeling capability to the domain of dynamic graphs by integrating the message passing mechanism of GNN into the evolution equations of neural ODE ( 22 ) . Despite these advances, real-world inter-node interaction patterns often exhibit significant diversity and time-varying nature. Specifically, diversity refers to the fact that inter-node interactions can take multiple types, and multiple different types of interactions may occur simultaneously at the same time; these interactions are not mutually exclusive but can coexist in parallel. Time-varying nature refers to the dynamic evolution of inter-node interaction patterns over time, manifested in the emergence and disappearance of interactions, switching among interaction types, and continuous changes in interaction strength. For example, in social networks, user interactions are diverse (e.g., likes, comments, shares) and show significant dynamic evolution characteristics ( 32 ) ; in infectious disease transmission networks, viruses exhibit multiple transmission modes (e.g., sexual transmission, bloodborne transmission, airborne transmission ( 30 ) ), which present significant time-dependent features across different stages ( 14 ; 40 ; 35 ) . Existing graph neural ODE models ( 13 ; 22 ) typically rely on a unified message passing mechanism, which assumes that all nodes share the same message passing function. While this assumption simplifies computation to some extent, a unified message passing function fails to capture the diversity and time-varying nature of inter-node interaction patterns, thereby limiting the model’s expressive power and interpretability ( 43 ) . Consequently, how to effectively model the diversity and time-varying nature of inter-node interaction patterns has emerged as a key challenge in dynamic graph learning. To address this challenge, we propose Time-varying Interaction Graph Ordinary Differential Equations (TI-ODE). The core idea of this model is to represent the evolution function in the graph ODE as a set of learnable interaction basis functions, each corresponding to a specific type of inter-node interaction, which are dynamically combined with time-dependent learnable weights. Importantly, the proposed formulation departs from conventional attention or reweighting schemes by representing interaction dynamics through a functional basis expansion, instead of relying on a single shared function to parameterize all pairwise relationships. The main contributions of this work are summarized as follows: • We propose a graph neural ODE model, TI-ODE, which models inter-node interactions in the graph ODE evolution function as a time-dependent combination of multiple basis functions, capturing both the diversity of interaction patterns and their time-varying nature. • We theoretically prove that TI-ODE exhibits higher robustness compared to models that adopt a unified message passing function. • Experimental results demonstrate that TI-ODE consistently outperforms existing state-of-the-art models across six diverse benchmarks, including two physical dynamics datasets, two molecular dynamics datasets, and two real-world datasets. To further validate its practical utility, we evaluate TI-ODE on the Covid dataset, demonstrating its superior interpretability and generalizability. Furthermore, additional empirical evaluations indicate that TI-ODE exhibits significantly enhanced robustness compared to models that rely on a unified message-passing function. 2 Related Work 2.1 GNN for Dynamic Graph Representation Owing to the inherent advantages of GNN in modeling complex structural relations ( 29 ) , they have been widely adopted in dynamic graph representation learning in recent years ( 44 ) . Early studies largely followed the paradigm of static GNN by stacking recurrent architectures over discrete time to capture temporal dependencies ( 44 ) , including DCRNN ( 20 ) , GraphWaveNet ( 39 ) , EvolveGCN ( 27 ) , WD-GCN ( 25 ) , AGCRN ( 1 ) , TTS-AMP ( 4 ) and T&S-AMP ( 4 ) . However, discrete-time models typically rely on a fixed sampling frequency, making them ill-suited for irregularly sampled or event-driven dynamic graphs. Consequently, continuous-time dynamic graph models have attracted increasing attention ( 44 ) , such as DyREP ( 33 ) , JODIE ( 17 ) , TGAT ( 41 ) , and TREND ( 38 ) . Although these methods support continuous-time timestamps, they often still depend on predefined time discretization and fixed update mechanisms, which can lead to information loss and temporal blurring ( 13 ) . In contrast, we

Read the complete text