Skip to content

Survey and presentation about Sum-Product Networks (SPNs)

License

Notifications You must be signed in to change notification settings

AlexPasqua/Sum-Product-Networks

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

1 Commit
 
 
 
 
 
 
 
 

Repository files navigation

Sum Product Networks (SPNs)

Survey and presentation about Sum-Product Networks (SPNs), created for the course Beyond Deep Learning: Selected Topics on Novel Challenges @ Technical University of Munich (TUM)

Abstract

Sum-Product Networks (SPNs) are a probabilistic framework that allows building tractable models from data, making it possible to perform various inference tasks in polynomial time. They are based on a rooted directed acyclic graph in which terminal nodes are the indicators of the input random variables (they can be seen as univariate probability distributions) and internal nodes divides themselves into sum nodes, which perform mixtures of sub-distributions, and product nodes, which perform factorizations. SPNs are related to probabilistic graphical models, such as Bayesian Networks, but the latter often suffer from tractability issues, since exact inference often requires exponential time. SPNs are also linked to Neural Networks (NNs), they share some similarities but there exist also many differences between these two classes of models. In some cases they can be used as alternatives to each other, to tackle similar kinds of problems, but they can also be applied in a complementary way, to exploit the best characteristics of both.

Note: the report follows the style guidelines of NeurIPS 2021

About

Survey and presentation about Sum-Product Networks (SPNs)

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published