A Recipe for Causal Graph Regression: Confounding Effects Revisited

TL;DR

This paper extends causal graph learning techniques from classification to regression tasks, addressing confounding effects and demonstrating improved out-of-distribution generalization in graph neural networks.

Contribution

It generalizes causal intervention methods to graph regression and introduces contrastive learning approaches for better confounder handling.

Findings

Effective causal interventions improve regression OOD performance.

Contrastive learning enhances confounder representation in regression.

Proposed methods outperform baselines on graph OOD benchmarks.

Abstract

Through recognizing causal subgraphs, causal graph learning (CGL) has risen to be a promising approach for improving the generalizability of graph neural networks under out-of-distribution (OOD) scenarios. However, the empirical successes of CGL techniques are mostly exemplified in classification settings, while regression tasks, a more challenging setting in graph learning, are overlooked. We thus devote this work to tackling causal graph regression (CGR); to this end we reshape the processing of confounding effects in existing CGL studies, which mainly deal with classification. Specifically, we reflect on the predictive power of confounders in graph-level regression, and generalize classification-specific causal intervention techniques to regression through a lens of contrastive learning. Extensive experiments on graph OOD benchmarks validate the efficacy of our proposals for CGR. The model implementation and the code are provided on https://github.com/causal-graph/CGR.