Preserving Spatial-Temporal Relationship with Adaptive Node Sampling in Hierarchical Dynamic Graph Transformers

Linh Thi Hoang, Tuan Dung Pham, Son T. Mai, Viet Cuong Ta
Proceedings of the 16th Asian Conference on Machine Learning, PMLR 260:1176-1191, 2025.

Abstract

Dynamic Graph Transformers (DGTs) have demonstrated remarkable performance in various applications, such as social networks, traffic forecasting, and recommendation systems. Despite their effectiveness in capturing long-range dependencies, training DGTs for large graphs remains a challenge. Mini-batch training is usually used to alleviate this challenge but this approach often fails to capture complex dependencies or sacrifice performance. To deal with the above problems, we propose the A_daptive Node S_ampling in H_ierarchical D_ynamic G_raph T_ransformers (ASH-DGT) architecture that focuses on sampling the set of suitable nodes preserving spatial-temporal relationships in the dynamic graph for training DGTs. Unlike previous methods that use random sampling or structural sampling, our motivation is that the contribution of nodes to learning performance can be time-sensitive, while we still care about spatial correlation in the dynamic graph with consideration to the global and local structure of the graph. Through extensive evaluations on popular real-world datasets for node classification and link prediction, ASH-DGT consistently outperforms multiple state-of-the-art methods, achieving both higher accuracy and significant improvements in training efficiency.

Cite this Paper

BibTeX
@InProceedings{pmlr-v260-hoang25a, title = {Preserving Spatial-Temporal Relationship with Adaptive Node Sampling in Hierarchical Dynamic Graph Transformers}, author = {Hoang, Linh Thi and Pham, Tuan Dung and Mai, Son T. and Ta, Viet Cuong}, booktitle = {Proceedings of the 16th Asian Conference on Machine Learning}, pages = {1176--1191}, year = {2025}, editor = {Nguyen, Vu and Lin, Hsuan-Tien}, volume = {260}, series = {Proceedings of Machine Learning Research}, month = {05--08 Dec}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v260/main/assets/hoang25a/hoang25a.pdf}, url = {https://proceedings.mlr.press/v260/hoang25a.html}, abstract = {Dynamic Graph Transformers (DGTs) have demonstrated remarkable performance in various applications, such as social networks, traffic forecasting, and recommendation systems. Despite their effectiveness in capturing long-range dependencies, training DGTs for large graphs remains a challenge. Mini-batch training is usually used to alleviate this challenge but this approach often fails to capture complex dependencies or sacrifice performance. To deal with the above problems, we propose the $\underline{A}$daptive Node $\underline{S}$ampling in $\underline{H}$ierarchical $\underline{D}$ynamic $\underline{G}$raph $\underline{T}$ransformers (ASH-DGT) architecture that focuses on sampling the set of suitable nodes preserving spatial-temporal relationships in the dynamic graph for training DGTs. Unlike previous methods that use random sampling or structural sampling, our motivation is that the contribution of nodes to learning performance can be time-sensitive, while we still care about spatial correlation in the dynamic graph with consideration to the global and local structure of the graph. Through extensive evaluations on popular real-world datasets for node classification and link prediction, ASH-DGT consistently outperforms multiple state-of-the-art methods, achieving both higher accuracy and significant improvements in training efficiency.} }
Endnote
%0 Conference Paper %T Preserving Spatial-Temporal Relationship with Adaptive Node Sampling in Hierarchical Dynamic Graph Transformers %A Linh Thi Hoang %A Tuan Dung Pham %A Son T. Mai %A Viet Cuong Ta %B Proceedings of the 16th Asian Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2025 %E Vu Nguyen %E Hsuan-Tien Lin %F pmlr-v260-hoang25a %I PMLR %P 1176--1191 %U https://proceedings.mlr.press/v260/hoang25a.html %V 260 %X Dynamic Graph Transformers (DGTs) have demonstrated remarkable performance in various applications, such as social networks, traffic forecasting, and recommendation systems. Despite their effectiveness in capturing long-range dependencies, training DGTs for large graphs remains a challenge. Mini-batch training is usually used to alleviate this challenge but this approach often fails to capture complex dependencies or sacrifice performance. To deal with the above problems, we propose the $\underline{A}$daptive Node $\underline{S}$ampling in $\underline{H}$ierarchical $\underline{D}$ynamic $\underline{G}$raph $\underline{T}$ransformers (ASH-DGT) architecture that focuses on sampling the set of suitable nodes preserving spatial-temporal relationships in the dynamic graph for training DGTs. Unlike previous methods that use random sampling or structural sampling, our motivation is that the contribution of nodes to learning performance can be time-sensitive, while we still care about spatial correlation in the dynamic graph with consideration to the global and local structure of the graph. Through extensive evaluations on popular real-world datasets for node classification and link prediction, ASH-DGT consistently outperforms multiple state-of-the-art methods, achieving both higher accuracy and significant improvements in training efficiency.
APA
Hoang, L.T., Pham, T.D., Mai, S.T. & Ta, V.C.. (2025). Preserving Spatial-Temporal Relationship with Adaptive Node Sampling in Hierarchical Dynamic Graph Transformers. Proceedings of the 16th Asian Conference on Machine Learning, in Proceedings of Machine Learning Research 260:1176-1191 Available from https://proceedings.mlr.press/v260/hoang25a.html.

Related Material