Rethinking the impact of noisy labels in graph classification: A utility and privacy perspective

De Li, Xianxian Li, Zeming Gan, Qiyu Li, Bin Qu, Jinyan Wang

Abstract: Highlights•We rethink the influence of noise labels on graph classification from the perspective of model utility and data privacy.•Two-stage noise sample selection is used to achieve accurate noise data separation.•Dual-view information is proposed to guide noise label correction.•Supervised graph contrast learning is introduced to reduce the ability of member inference attacks.