Cluster-Dags as Powerful Background Knowledge For Causal Discovery

2025-12-10

AI Summary

Learning causal graphs from high-dimensional, complex data faces challenges of insufficient prior knowledge and poor scalability. To address this, we propose Cluster-DAGs—a structured DAG prior framework based on variable clustering—that offers both flexibility and interpretability, outperforming conventional hierarchical priors. Building upon this, we design two novel algorithms: Cluster-PC for fully observed settings and Cluster-FCI for partially observed settings with latent variables or selection bias. Both integrate constraint-based causal discovery, conditional independence testing, and clustering-driven variable grouping. In extensive simulations, Cluster-PC and Cluster-FCI significantly outperform standard PC and FCI baselines, achieving substantial improvements in accuracy and robustness of causal structure recovery.

Technology Category

• Reasoning under Uncertainty: Causality
• Machine Learning: Causal Learning
• Constraint Satisfaction and Optimization: Other Foundations of Constraint Satisfaction

Application Category

• Graph Algorithms and Modeling for the Web: Algorithms and analysis for incomplete, noisy, or partially observed Web-related graphs
• Semantics and Knowledge: Methods, algorithms and applications for the development of semantic models, knowledge graphs and other forms of structured data models with machine-interpretable semantics
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Finding cause-effect relationships is of key importance in science. Causal discovery aims to recover a graph from data that succinctly describes these cause-effect relationships. However, current methods face several challenges, especially when dealing with high-dimensional data and complex dependencies. Incorporating prior knowledge about the system can aid causal discovery. In this work, we leverage Cluster-DAGs as a prior knowledge framework to warm-start causal discovery. We show that Cluster-DAGs offer greater flexibility than existing approaches based on tiered background knowledge and introduce two modified constraint-based algorithms, Cluster-PC and Cluster-FCI, for causal discovery in the fully and partially observed setting, respectively. Empirical evaluation on simulated data demonstrates that Cluster-PC and Cluster-FCI outperform their respective baselines without prior knowledge.

Problem

• Incorporating Cluster-DAGs as flexible prior knowledge to improve causal discovery
• Addressing challenges in high-dimensional data and complex dependencies for causal graph recovery
• Enhancing constraint-based algorithms with Cluster-PC and Cluster-FCI for better performance

Innovation

• Using Cluster-DAGs as flexible prior knowledge framework
• Introducing modified algorithms Cluster-PC and Cluster-FCI
• Enhancing causal discovery in fully and partially observed settings