Kernel entropy graph isomorphism network for graph classification

doi:10.1016/j.patcog.2026.113182

Kernel entropy graph isomorphism network for graph classification

2026-01-28

0

PRE

AI

DOI：10.1016/j.patcog.2026.113182

Original

Original request for help

Save

Abstract

En 中文

• We propose a fusion method for graph neural networks and graph kernels based on the attention mechanism. • We use a shortest path matching kernel to extract global structural features of graph. • We adopt an entropy-based graph isomorphism network to extract local structural features of graph. • We utilize the Nystom method kernel matrix for reduced-rank decomposition, which reduces the time consumed for model computation.

Kernel entropy graph isomorphism network for graph classification

Abstract

Journal

Organization