Unraveling the Impact of Heterophilic Structures on Graph Positive-Unlabeled Learning

TL;DR

This paper addresses the challenge of positive-unlabeled learning on graph data by introducing GPL, a method that reduces heterophily to improve label inference and classifier performance.

Contribution

The paper proposes GPL, a novel bilevel optimization approach that mitigates heterophily effects in graph PU learning, enhancing label estimation and classification accuracy.

Findings

GPL significantly outperforms baseline methods across multiple datasets.

Heterophily reduction improves class-prior estimation and label inference.

The bilevel optimization framework effectively addresses heterophilic structures in graphs.

Abstract

While Positive-Unlabeled (PU) learning is vital in many real-world scenarios, its application to graph data still remains under-explored. We unveil that a critical challenge for PU learning on graph lies on the edge heterophily, which directly violates the irreducibility assumption for Class-Prior Estimation (class prior is essential for building PU learning algorithms) and degenerates the latent label inference on unlabeled nodes during classifier training. In response to this challenge, we introduce a new method, named Graph PU Learning with Label Propagation Loss (GPL). Specifically, GPL considers learning from PU nodes along with an intermediate heterophily reduction, which helps mitigate the negative impact of the heterophilic structure. We formulate this procedure as a bilevel optimization that reduces heterophily in the inner loop and efficiently learns a classifier in the outer loop. Extensive experiments across a variety of datasets have shown that GPL significantly outperforms baseline methods, confirming its effectiveness and superiority.