IBPL: Information Bottleneck-based Prompt Learning for graph out-of-distribution detection

Yanan Cao, Fengzhao Shi, Qing Yu, Xixun Lin, Chuan Zhou, Lixin Zou, Peng Zhang, Zhao Li, Dawei Yin

Abstract: Highlights•We study how to eliminate the overlapping features to enlarge the score gap.•We propose the IB-based graph prompt to optimize node features and graph structure.•Our method yields the superior results over existing strong approaches empirically.