Structured Entropy Quantification for uncertainty-aware graph learning

doi:10.1016/j.patcog.2026.114794

Structured Entropy Quantification for uncertainty-aware graph learning

2026-09-03

0

PRE

AI

DOI：10.1016/j.patcog.2026.114794

Original

Original request for help

Save

Abstract

En 中文

• Dual-entropy quantifies edge reliability from local and global views. • Asymmetric conditional entropy models directional causal flows. • Closed-loop calibration adapts weights via gradient feedback. • Plug-and-play module enhances any GNN backbone.

Keywords:

Graph neural network

Entropy quantification

Uncertainty modeling

Information bottleneck

Structured Entropy Quantification for uncertainty-aware graph learning

Abstract

Journal

Organization