Structured Factored Inference for Probabilistic Programming

Avi Pfeffer, Brian Ruttenberg, William Kretschmer, Alison OConnor
Proceedings of the Twenty-First International Conference on Artificial Intelligence and Statistics, PMLR 84:1224-1232, 2018.

Abstract

Probabilistic reasoning on complex real-world models is computationally challenging. Inference algorithms have been developed that work well on specific models or on parts of general models, but they require significant hand-engineering to apply to full-scale problems. Probabilistic programming (PP) enables the expression of rich probabilistic models, but inference remains a bottleneck in many applications. Factored inference is one of the main approaches to inference in graphical models, but has trouble scaling up to some hard problems expressible as probabilistic programs. We present structured factored inference (SFI), a framework that enables factored inference algorithms to scale to significantly more complex programs. Using models encoded in a PP language, SFI provides a sound means to decompose a model into submodels, apply an algorithm to each submodel, and combine results to answer a query. Our results show that SFI successfully reasons on models where standard factored inference algorithms fail due to computational complexity. SFI is nearly as accurate as exact inference and is as fast as approximate inference methods.

Cite this Paper

BibTeX
@InProceedings{pmlr-v84-pfeffer18a, title = {Structured Factored Inference for Probabilistic Programming}, author = {Pfeffer, Avi and Ruttenberg, Brian and Kretschmer, William and OConnor, Alison}, booktitle = {Proceedings of the Twenty-First International Conference on Artificial Intelligence and Statistics}, pages = {1224--1232}, year = {2018}, editor = {Storkey, Amos and Perez-Cruz, Fernando}, volume = {84}, series = {Proceedings of Machine Learning Research}, month = {09--11 Apr}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v84/pfeffer18a/pfeffer18a.pdf}, url = {https://proceedings.mlr.press/v84/pfeffer18a.html}, abstract = {Probabilistic reasoning on complex real-world models is computationally challenging. Inference algorithms have been developed that work well on specific models or on parts of general models, but they require significant hand-engineering to apply to full-scale problems. Probabilistic programming (PP) enables the expression of rich probabilistic models, but inference remains a bottleneck in many applications. Factored inference is one of the main approaches to inference in graphical models, but has trouble scaling up to some hard problems expressible as probabilistic programs. We present structured factored inference (SFI), a framework that enables factored inference algorithms to scale to significantly more complex programs. Using models encoded in a PP language, SFI provides a sound means to decompose a model into submodels, apply an algorithm to each submodel, and combine results to answer a query. Our results show that SFI successfully reasons on models where standard factored inference algorithms fail due to computational complexity. SFI is nearly as accurate as exact inference and is as fast as approximate inference methods.} }
Endnote
%0 Conference Paper %T Structured Factored Inference for Probabilistic Programming %A Avi Pfeffer %A Brian Ruttenberg %A William Kretschmer %A Alison OConnor %B Proceedings of the Twenty-First International Conference on Artificial Intelligence and Statistics %C Proceedings of Machine Learning Research %D 2018 %E Amos Storkey %E Fernando Perez-Cruz %F pmlr-v84-pfeffer18a %I PMLR %P 1224--1232 %U https://proceedings.mlr.press/v84/pfeffer18a.html %V 84 %X Probabilistic reasoning on complex real-world models is computationally challenging. Inference algorithms have been developed that work well on specific models or on parts of general models, but they require significant hand-engineering to apply to full-scale problems. Probabilistic programming (PP) enables the expression of rich probabilistic models, but inference remains a bottleneck in many applications. Factored inference is one of the main approaches to inference in graphical models, but has trouble scaling up to some hard problems expressible as probabilistic programs. We present structured factored inference (SFI), a framework that enables factored inference algorithms to scale to significantly more complex programs. Using models encoded in a PP language, SFI provides a sound means to decompose a model into submodels, apply an algorithm to each submodel, and combine results to answer a query. Our results show that SFI successfully reasons on models where standard factored inference algorithms fail due to computational complexity. SFI is nearly as accurate as exact inference and is as fast as approximate inference methods.
APA
Pfeffer, A., Ruttenberg, B., Kretschmer, W. & OConnor, A.. (2018). Structured Factored Inference for Probabilistic Programming. Proceedings of the Twenty-First International Conference on Artificial Intelligence and Statistics, in Proceedings of Machine Learning Research 84:1224-1232 Available from https://proceedings.mlr.press/v84/pfeffer18a.html.

Related Material