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

02/20/2021
by   Matej Zečević, et al.
16

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.

READ FULL TEXT

page 1

page 7

page 8

page 12

page 13

page 14

research
10/22/2021

On the Tractability of Neural Causal Inference

Roth (1996) proved that any form of marginal inference with probabilisti...
research
01/29/2020

Interventions and Counterfactuals in Tractable Probabilistic Models: Limitations of Contemporary Transformations

In recent years, there has been an increasing interest in studying causa...
research
10/19/2021

Explaining Deep Tractable Probabilistic Models: The sum-product network case

We consider the problem of explaining a tractable deep probabilistic mod...
research
01/29/2020

On Constraint Definability in Tractable Probabilistic Models

Incorporating constraints is a major concern in probabilistic machine le...
research
11/11/2016

The Sum-Product Theorem: A Foundation for Learning Tractable Models

Inference in expressive probabilistic models is generally intractable, w...
research
02/07/2022

Causal Inference Using Tractable Circuits

The aim of this paper is to discuss a recent result which shows that pro...
research
06/30/2017

Probabilistic Active Learning of Functions in Structural Causal Models

We consider the problem of learning the functions computing children fro...

Please sign up or login with your details

Forgot password? Click here to reset