Aggregation-aware MLP: An unsupervised approach for graph message-passing

doi:10.1016/j.patcog.2026.113877

Aggregation-aware MLP: An unsupervised approach for graph message-passing

2026-04-29

0

PRE

AI

DOI：10.1016/j.patcog.2026.113877

Original

Original request for help

Save

Abstract

En 中文

• Shift the focus from GNN’s aggregation to make representations aggregation-adaptive. • Propose an unsupervised framework that unifies homophilic and heterophilic graphs. • In theoretical analysis, extend traditional Grouping Effect to a high- order version.

Keywords:

Graph Neural Networks

Message Passing

Aggregation-adaptive

Unsupervised Learning

Homophilic Graphs

Aggregation-aware MLP: An unsupervised approach for graph message-passing

Abstract

Journal

Organization