Deep Causal Graph Clustering Network for Functional Connectivity Data

Abstract

Deep clustering plays a crucial role in uncovering functional brain structures in cognitive neuroscience. However, existing approaches often learn redundant representations irrelevant to clustering and fail to capture the causal dependencies inherent in brain functional connectivity data. To address these challenges, we propose Deep Causal Graph Clustering (DCGC), a general framework that jointly learns causal representations and performs clustering in a unified manner. Specifically, DCGC employs a self-representation module to uncover causal relations between latent features and cluster assignments. In addition, a causal regularization objective constrains the latent space to eliminate redundancy based on the learned causal graph. Furthermore, we provide theoretical guarantees for the joint causal–clustering optimization by deriving an explicit upper bound for the objective function and establishing convergence properties. Experiments on eight publicly available functional connectivity datasets demonstrate that DCGC consistently outperforms state-of-the-art deep clustering methods, offering a principled causal perspective for unsupervised brain network analysis.

Keywords

Causal clustering / Graph Network / brain network analysis

RIGHTS & PERMISSIONS

Higher Education Press 2026

Just Accepted

This article has successfully passed peer review and final editorial review, and will soon enter typesetting, proofreading and other publishing processes. The currently displayed version is the accepted final manuscript. The officially published version will be updated with format, DOI and citation information upon launch. We recommend that you pay attention to subsequent journal notifications and preferentially cite the officially published version. Thank you for your support and cooperation.