Causal Graph Learning via Distributional Invariance of Cause-Effect Relationship

TL;DR

This paper proposes a novel causal graph learning method based on the invariance of effect distributions conditioned on causes across different prior cause distributions, enabling efficient and scalable causal discovery from observational data.

Contribution

It introduces a distributional invariance-based framework and an efficient algorithm for causal graph recovery with improved scalability and competitive accuracy.

Findings

Outperforms existing methods in scalability, reducing processing time by up to 25x.

Achieves superior or comparable accuracy on large-scale benchmark datasets.

Utilizes invariance of effect distributions to identify causal relationships effectively.

Abstract

This paper introduces a new framework for recovering causal graphs from observational data, leveraging the observation that the distribution of an effect, conditioned on its causes, remains invariant to changes in the prior distribution of those causes. This insight enables a direct test for potential causal relationships by checking the variance of their corresponding effect-cause conditional distributions across multiple downsampled subsets of the data. These subsets are selected to reflect different prior cause distributions, while preserving the effect-cause conditional relationships. Using this invariance test and exploiting an (empirical) sparsity of most causal graphs, we develop an algorithm that efficiently uncovers causal relationships with quadratic complexity in the number of observational variables, reducing the processing time by up to 25x compared to state-of-the-art methods. Our empirical experiments on a varied benchmark of large-scale datasets show superior or equivalent performance compared to existing works, while achieving enhanced scalability.