Interventional Sum-Product Networks: Causal Inference with Tractable Probabilistic Models

Abstract

While probabilistic models are an important tool for studying causality, doing so suffers from the intractability of inference. As a step towards tractable causal models, we consider the problem of learning interventional distributions using sum-product networks (SPNs) that are over-parameterized by gate functions, e.g., neural networks. Providing an arbitrarily intervened causal graph as input, effectively subsuming Pearl's do-operator, the gate function predicts the parameters of the SPN. The resulting interventional SPNs are motivated and illustrated by a structural causal model themed around personal health. Our empirical evaluation on three benchmark data sets as well as a synthetic health data set clearly demonstrates that interventional SPNs indeed are both expressive in modelling and flexible in adapting to the interventions.

arXiv commentMain paper: 10 pages, References: 3 pages, Appendix: 8 pages. Main paper: 6 figures, Appendix: 5 figures

References in corpus (11)

• MADE: Masked Autoencoder for Distribution Estimation
• Causality for Machine Learning
• Causal Reasoning from Meta-reinforcement Learning
• Counterfactual Off-Policy Evaluation with Gumbel-Max Structural Causal Models
• CausalML: Python Package for Causal Machine Learning
• Learning Neural Causal Models from Unknown Interventions
• Learning and Testing Causal Models with Interventions
• DoWhy: An End-to-End Library for Causal Inference
• Conditional Sum-Product Networks: Imposing Structure on Deep Probabilistic Architectures
• Sum-product networks: A survey
• Interventions and Counterfactuals in Tractable Probabilistic Models: Limitations of Contemporary Transformations