RECLAIM: Cyclic Causal Discovery Amid Measurement Noise

TL;DR

RECLAIM is a novel causal discovery framework that effectively uncovers cyclic causal structures from noisy measurements using EM and normalizing flows, with proven theoretical guarantees and validated on real data.

Contribution

It introduces a method that handles cycles and measurement noise in causal discovery, with consistency guarantees and practical validation.

Findings

Successfully uncovers cyclic causal structures in synthetic data.

Demonstrates effectiveness on real-world protein signaling datasets.

Provides theoretical guarantees for the proposed models.

Abstract

Uncovering causal relationships is a fundamental problem across science and engineering. However, most existing causal discovery methods assume acyclicity and direct access to the system variables -- assumptions that fail to hold in many real-world settings. For instance, in genomics, cyclic regulatory networks are common, and measurements are often corrupted by instrumental noise. To address these challenges, we propose RECLAIM, a causal discovery framework that natively handles both cycles and measurement noise. RECLAIM learns the causal graph structure by maximizing the likelihood of the observed measurements via expectation-maximization (EM), using residual normalizing flows for tractable likelihood computation. We consider two measurement models: (i) Gaussian additive noise, and (ii) a linear measurement system with additive Gaussian noise. We provide theoretical consistency guarantees for both the settings. Experiments on synthetic data and real-world protein signaling datasets demonstrate the efficacy of the proposed method.