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.

Major Problems in Clinical Psychological Science and How to Address them. Introducing a Multimodal Dynamical Network Approach

“…The EEMM applies evolutionary concepts of context-appropriate variation, selection, and retention to relevant biopsychosocial processes, such as cognition, affect, or physiology. It also emphasizes connections and al., 2023) from multimodal sources with state-of-the-art network analysis and causal inference methods (e.g., Shi et al, 2021;Zečević et al, 2021). MMDNs aim to integrate psychological processes (e.g., attention, expectations; e.g., Heeren, Jones et al, 2018b;Isvoranu et al, 2017;McNally et al, 2022), psychopathological symptoms (e.g., fear, listlessness, sleep disturbance; e.g., Barthel et al, 2020;Curtiss et al, 2018;Uhlhaas & Singer, 2015), and -whenever promising -neurobiological correlates (e.g., resting-state markers, neuromodulation, neurotransmitter systems; e.g., Cole et al, 2020;Kaiser et al, 2022;Kringelbach & Deco, 2020), providing a comprehensive, multimodal perspective on mental disorders (e.g., Taschereau-Dumouchel et al, 2022).…”