Learning Neural Causal Models from Unknown Interventions

TL;DR

This paper introduces a neural network-based continuous optimization framework for learning causal Bayesian network structures from both observational and interventional data, even when intervention targets are unknown.

Contribution

It presents a novel method that combines observational and interventional data using neural networks, capable of handling unknown intervention targets, advancing causal structure learning.

Findings

Achieves strong benchmark results on synthetic and real Bayesian networks.

Effective in recovering causal structures from mixed data types.

Handles unknown intervention targets successfully.

Abstract

Promising results have driven a recent surge of interest in continuous optimization methods for Bayesian network structure learning from observational data. However, there are theoretical limitations on the identifiability of underlying structures obtained from observational data alone. Interventional data provides much richer information about the underlying data-generating process. However, the extension and application of methods designed for observational data to include interventions is not straightforward and remains an open problem. In this paper we provide a general framework based on continuous optimization and neural networks to create models for the combination of observational and interventional data. The proposed method is even applicable in the challenging and realistic case that the identity of the intervened upon variable is unknown. We examine the proposed method in the setting of graph recovery both de novo and from a partially-known edge set. We establish strong benchmark results on several structure learning tasks, including structure recovery of both synthetic graphs as well as standard graphs from the Bayesian Network Repository.