Trust Your $\nabla$: Gradient-based Intervention Targeting for Causal Discovery

TL;DR

This paper introduces GIT, a gradient-based method for selecting optimal interventions in causal discovery, which is especially effective when data is limited, improving the efficiency of causal structure learning.

Contribution

The paper proposes a novel Gradient-based Intervention Targeting (GIT) method that leverages gradient estimators to identify informative interventions for causal discovery.

Findings

GIT performs comparably to existing methods in general cases.

GIT outperforms baselines in low-data regimes.

Extensive experiments validate GIT's effectiveness on simulated and real datasets.

Abstract

Inferring causal structure from data is a challenging task of fundamental importance in science. Observational data are often insufficient to identify a system's causal structure uniquely. While conducting interventions (i.e., experiments) can improve the identifiability, such samples are usually challenging and expensive to obtain. Hence, experimental design approaches for causal discovery aim to minimize the number of interventions by estimating the most informative intervention target. In this work, we propose a novel Gradient-based Intervention Targeting method, abbreviated GIT, that 'trusts' the gradient estimator of a gradient-based causal discovery framework to provide signals for the intervention acquisition function. We provide extensive experiments in simulated and real-world datasets and demonstrate that GIT performs on par with competitive baselines, surpassing them in the low-data regime.