Cauchy Graph Convolutional Networks

Taurai Muvunza, Yang Li, Ercan Engin Kuruoglu

Abstract: Highlights•We introduce Cauchy Graphical Models (CGM) that can be represented as directed acyclic graphs (DAGs) to model impulsive noise in random variables.•We propose Minimum Dispersion Criterion (MDC), a score-based DAG selection method for optimal CGM.•We present Cauchy GCN which leverages CGM-learned graphs to boost GCN classification performance.•We conduct an extensive experimental campaign to validate the efficacy of our approach.

External IDs:dblp:journals/ijar/MuvunzaLK25