Augmentation-free dynamic graph contrastive learning based on transformers

doi:10.1016/j.eswa.2025.129028

Augmentation-free dynamic graph contrastive learning based on transformers

2025-07-15

0

PRE

AI

DOI：10.1016/j.eswa.2025.129028

Original

Original request for help

Save

Abstract

En 中文

• Avoid data degradation with augmentation-free dynamic graph contrastive learning. • Dynamically adjust Transformer patch sizes to capture multi-scale temporal features. • Select positive samples using nearest-neighbor interaction frequency optimization. • Optimize node representation and link prediction in an end-to-end learning framework. • Achieve better performance than state-of-the-art methods, even with noisy data.

Augmentation-free dynamic graph contrastive learning based on transformers

Abstract

Journal

Organization