Masked Heterogeneous Graph Attention Network for robust recommendation

doi:10.1016/j.asoc.2025.113596

Masked Heterogeneous Graph Attention Network for robust recommendation

2025-07-23

0

PRE

AI

DOI：10.1016/j.asoc.2025.113596

Original

Original request for help

Save

Abstract

En 中文

• Investigated HGNN vulnerabilities to attack diffusion and attention weight inertia. • Proposed MHGAN using propagation constraints and masking to enhance robustness. • MHGAN validated via experiments on three datasets under adversarial attack scenarios.

Keywords:

HGNN

adversarial attack

robustness

propagation constraints

masking

Masked Heterogeneous Graph Attention Network for robust recommendation

Abstract

Journal

Organization