Sum-product networks: A survey

Paris, Iago; Sánchez-Cauce, Raquel; Díez, Francisco Javier Herrero

doi:10.48550/arxiv.2004.01167

Cited by 2 publications

(2 citation statements)

References 42 publications

(71 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Several different types of SPNs have also been studied such as Random SPN [Peharz et al, 2020b], Credal SPNs [Levray and Belle, 2020] and Sum-Product-Quotient Networks [Sharir and Shashua, 2018]) to name a few. For more details readers are referred to the survey of París, Sánchez-Cauce, and Díez [2020].…”

Section: Background and Related Workmentioning

confidence: 99%

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

Zečević¹,

Dhami²,

Karanam³

et al. 2021

Preprint

View full text Add to dashboard Cite

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.

show abstract

Section: Background and Related Workmentioning

confidence: 99%

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

Zečević¹,

Dhami²,

Karanam³

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

“…They are more appliable in lots of tasks such as online causal inference (Pearl et al, 2009) and query selectivity estimation (Getoor & Taskar, 2001), which have strict requirements on both inference accuracy and speed of the deployed models. Therefore, PGMs have recently re-attracted considerable research interests in the ML community (Zheng et al, 2018;Scanagatta et al, 2019;París et al, 2020). Much efforts (Rooshenas & Lowd, 2014;Vergari et al, 2015;Desana & Schnörr, 2020;Rahman & Gogate, 2016;Shao et al, 2019;Sharir & Shashua, 2018;Rahman et al, 2014;Darwiche, 2009;Boutilier et al, 2013) have been devoted to improving the accuracy and tractability (inference speed) of PGMs.…”

Section: Introductionmentioning

confidence: 99%

FSPN: A New Class of Probabilistic Graphical Model

Wu,

Zhu,

Pfadler

et al. 2020

Preprint

View full text Add to dashboard Cite

We introduce factorize-sum-split-product networks (FSPNs), a new class of probabilistic graphical models (PGMs). FSPNs are designed to overcome the drawbacks of existing PGMs in terms of estimation accuracy and inference efficiency. Specifically, Bayesian networks (BNs) have low inference speed and performance of tree-structured sum-product networks(SPNs) significantly degrades in presence of highly correlated variables. FSPNs absorb their advantages by adaptively modeling the joint distribution of variables according to their dependence degree, so that one can simultaneously attain the two desirable goals-high estimation accuracy and fast inference speed. We present efficient probability inference and structure learning algorithms for FSPNs, along with a theoretical analysis and extensive evaluation evidence. Our experimental results on synthetic and benchmark datasets indicate the superiority of FSPN over other PGMs.

show abstract

Sum-product networks: A survey

Cited by 2 publications

References 42 publications

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

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

FSPN: A New Class of Probabilistic Graphical Model

Contact Info

Product

Resources

About