Entropic Causal Inference: Graph Identifiability

2025-09-19

AI Summary

This paper addresses the problem of identifiably learning causal graph structures from observational data. For the bivariate case, it establishes, for the first time, identifiability of causal direction via entropy minimization under relaxed assumptions—dispensing with conventional requirements such as additive noise models. Extending this insight, the paper develops the first entropy-based identifiability theory for general multivariate causal graphs. It further proposes a sequential pruning algorithm that integrates information-theoretic ancestral relation testing with heuristic search to efficiently reconstruct arbitrary causal structures. The method demonstrates significant performance gains over state-of-the-art baselines across diverse synthetic benchmarks and validates its effectiveness and robustness on real-world datasets.

Technology Category

• Machine Learning: Causal Learning
• Reasoning under Uncertainty: Causality
• Knowledge Representation and Reasoning: Action, Change, and Causality

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

Entropic causal inference is a recent framework for learning the causal graph between two variables from observational data by finding the information-theoretically simplest structural explanation of the data, i.e., the model with smallest entropy. In our work, we first extend the causal graph identifiability result in the two-variable setting under relaxed assumptions. We then show the first identifiability result using the entropic approach for learning causal graphs with more than two nodes. Our approach utilizes the property that ancestrality between a source node and its descendants can be determined using the bivariate entropic tests. We provide a sound sequential peeling algorithm for general graphs that relies on this property. We also propose a heuristic algorithm for small graphs that shows strong empirical performance. We rigorously evaluate the performance of our algorithms on synthetic data generated from a variety of models, observing improvement over prior work. Finally we test our algorithms on real-world datasets.

Problem

• Extending causal graph identifiability under relaxed two-variable assumptions
• Establishing first entropic identifiability for graphs with multiple nodes
• Developing algorithms to determine ancestral relationships using bivariate tests

Innovation

• Extends causal identifiability under relaxed assumptions
• Uses bivariate entropic tests for ancestrality determination
• Proposes sequential peeling algorithm for general graphs