Filters








1,222 Hits in 7.6 sec

Gen: a general-purpose probabilistic programming system with programmable inference

Marco F. Cusumano-Towner, Feras A. Saad, Alexander K. Lew, Vikash K. Mansinghka
2019 Proceedings of the 40th ACM SIGPLAN Conference on Programming Language Design and Implementation - PLDI 2019  
This paper introduces Gen, a new probabilistic programming system with novel language constructs for modeling and for end-user customization and optimization of inference.  ...  Key Challenges Two key challenges in designing a practical general-purpose probabilistic programming system are: (i) achieving good performance for heterogeneous probabilistic models that combine black  ...  A Flexible Architecture for Modeling and Inference In existing probabilistic programming systems [13, 15, 48] , inference algorithm implementations are intertwined with the implementation of compilers  ... 
doi:10.1145/3314221.3314642 dblp:conf/pldi/Cusumano-Towner19 fatcat:bmjwmh7jhjf33gg6dywrdjop7y

Trace types and denotational semantics for sound programmable inference in probabilistic languages

Alexander K. Lew, Marco F. Cusumano-Towner, Benjamin Sherman, Michael Carbin, Vikash K. Mansinghka
2019 Proceedings of the ACM on Programming Languages (PACMPL)  
ACKNOWLEDGMENTS This material is based upon work supported by philanthropic gifts from the Siegel Family Foundation and from the Aphorism Foundation, and also by a research contract from the Intel Probabilistic  ...  a probabilistic program could be useful for other purposes, too.  ...  the simple, general-purpose variational families used by some probabilistic programming languages without programmable inference [Carpenter et al. 2017; Ge et al. 2018; Wingate and Weber 2013] .  ... 
doi:10.1145/3371087 fatcat:kf4vfvzt7rdvzencdtksbk3moq

Bayesian causal inference via probabilistic program synthesis [article]

Sam Witty, Alexander Lew, David Jensen, Vikash Mansinghka
2019 arXiv   pre-print
This approach also enables the use of general-purpose inference machinery for probabilistic programs to infer probable causal structures and parameters from data.  ...  This abstract describes a prototype of this approach in the Gen probabilistic programming language.  ...  Then, we use a more expressive probabilistic programming language, Gen, to encode a prior and likelihood over MiniStan programs, and to do inference.  ... 
arXiv:1910.14124v1 fatcat:ywbfuyq4cjfzhjrpvqf4tqjhbi

Optimization Framework with Minimum Description Length Principle for Probabilistic Programming [chapter]

Alexey Potapov, Vita Batishcheva, Sergey Rodionov
2015 Lecture Notes in Computer Science  
Application of the Minimum Description Length principle to optimization queries in probabilistic programming was investigated on the example of the C++ probabilistic programming library under development  ...  It was shown that incorporation of this criterion is essential for optimization queries to behave similarly to more common queries performing sampling in accordance with posterior distributions and automatically  ...  It can also be seen that general C++ code can be easily used together with our probabilistic programming library.  ... 
doi:10.1007/978-3-319-21365-1_34 fatcat:c66a32noyzhavpvzka36lit6em

Learning Proposals for Probabilistic Programs with Inference Combinators [article]

Sam Stites, Heiko Zimmermann, Hao Wu, Eli Sennesh, Jan-Willem van de Meent
2021 arXiv   pre-print
We develop operators for construction of proposals in probabilistic programs, which we refer to as inference combinators.  ...  The result is a framework for user-programmable variational methods that are correct by construction and can be tailored to specific models.  ...  Cusumano-Towner, Marco F., Saad, Feras A., Lew, Alexan- der K., Mansinghka, Vikash K. "Gen: A General-Purpose Probabilistic Programming System with Programmable Inference."  ... 
arXiv:2103.00668v3 fatcat:wc5yn2njabbzpeayykqk55cymu

Program Analysis of Probabilistic Programs [article]

Maria I. Gorinova
2022 arXiv   pre-print
No single inference algorithm can be used as a probabilistic programming back-end that is simultaneously reliable, efficient, black-box, and general.  ...  Probabilistic programming is a growing area that strives to make statistical analysis more accessible, by separating probabilistic modelling from probabilistic inference.  ...  Gen Gen (Cusumano-Towner et al., 2019 ) is a Julia-based PPL centred around the idea of programmable inference .  ... 
arXiv:2204.06868v1 fatcat:2dbonwruuzaopil4aijdeuz4mi

Functional programming for modular Bayesian inference

Adam Ścibior, Ohad Kammar, Zoubin Ghahramani
2018 Proceedings of the ACM on Programming Languages  
Probabilistic programs give us a formalism for expressing distributions by programming with the abstractions of Bayesian statistics: sampling from the prior distribution and changing the likelihood, also  ...  We evaluate our implementation against existing probabilistic programming systems and find it is already competitively performant, although we conjecture that existing functional programming optimisation  ...  We thank David Tolpin for providing Anglican and WebPPL scripts we used for benchmarking and Jeremy Yallop for his help with using the OCaml type system.  ... 
doi:10.1145/3236778 dblp:journals/pacmpl/ScibiorKG18 fatcat:g6tspxtoh5gfvho4iy6mdxu5ay

Modular Probabilistic Models via Algebraic Effects [article]

Minh Nguyen, Roly Perera, Meng Wang, Nicolas Wu
2022 arXiv   pre-print
Probabilistic programming languages (PPLs) allow programmers to construct statistical models and then simulate data or perform inference over them.  ...  Many PPLs restrict models to a particular instance of simulation or inference, limiting their reusability. In other PPLs, models are not readily composable.  ...  We also thank Alessio Zakaria who has been a constant source of support and drive behind our ideas, and the members of the Bristol Programming Languages research group for creating the wonderful environment  ... 
arXiv:2203.04608v3 fatcat:7nesx5lhvrbdtfe266vbqarv4e

Efficient synthesis of probabilistic programs

Aditya V. Nori, Sherjil Ozair, Sriram K. Rajamani, Deepak Vijaykeerthy
2015 Proceedings of the 36th ACM SIGPLAN Conference on Programming Language Design and Implementation - PLDI 2015  
A core difficulty in synthesizing probabilistic programs is computing the likelihood L(P | D) of a candidate program P generating data D.  ...  Our algorithm efficiently synthesizes a probabilistic program that is most consistent with the data.  ...  Acknowledgments We thank Timon Gehr and Martin Vechev for their help with fixing errors and improving the presentation of this paper.  ... 
doi:10.1145/2737924.2737982 dblp:conf/pldi/NoriORV15 fatcat:6k3nkkbzfra5vplekzxot3rsoi

Modular Probabilistic Models via Algebraic Effects [article]

Minh Nguyen
2022 Zenodo  
Artifact for "Modular Probabilistic Models via Algebraic Effects"  ...  We also thank Alessio Zakaria who has been a constant source of support and drive behind our ideas, and the members of the Bristol Programming Languages research group for creating the wonderful environment  ...  By integrating such notions into general-purpose languages, probabilistic programming languages (PPLs) allow programmers to build and execute probabilistic models.  ... 
doi:10.5281/zenodo.6651944 fatcat:6xku2qgnkzdzfg3jnozm5ser44

Modular Probabilistic Models via Algebraic Effects [article]

Minh Nguyen
2022 Zenodo  
Artifact for "Modular Probabilistic Models via Algebraic Effects"  ...  We also thank Alessio Zakaria who has been a constant source of support and drive behind our ideas, and the members of the Bristol Programming Languages research group for creating the wonderful environment  ...  By integrating such notions into general-purpose languages, probabilistic programming languages (PPLs) allow programmers to build and execute probabilistic models.  ... 
doi:10.5281/zenodo.6814316 fatcat:4tjiabd6bba5xapliaks7p7q54

Efficient synthesis of probabilistic programs

Aditya V. Nori, Sherjil Ozair, Sriram K. Rajamani, Deepak Vijaykeerthy
2015 SIGPLAN notices  
A core difficulty in synthesizing probabilistic programs is computing the likelihood L(P | D) of a candidate program P generating data D.  ...  Our algorithm efficiently synthesizes a probabilistic program that is most consistent with the data.  ...  Acknowledgments We thank Timon Gehr and Martin Vechev for their help with fixing errors and improving the presentation of this paper.  ... 
doi:10.1145/2813885.2737982 fatcat:4zrdv2tnhvb67dowwxnpflj6ii

Modular Probabilistic Models via Algebraic Effects [article]

Minh Nguyen
2022 Zenodo  
Artifact for "Modular Probabilistic Models via Algebraic Effects"  ...  We also thank Alessio Zakaria who has been a constant source of support and drive behind our ideas, and the members of the Bristol Programming Languages research group for creating the wonderful environment  ...  By integrating such notions into general-purpose languages, probabilistic programming languages (PPLs) allow programmers to build and execute probabilistic models.  ... 
doi:10.5281/zenodo.6651962 fatcat:zw2uepabivdkbamlbmgcfvhwia

Static Analysis for Probabilistic Programs [article]

Ryan Bernstein
2019 arXiv   pre-print
This field of static analysis for probabilistic programming (SAPP) is young and unorganized, consisting of a constellation of techniques with various goals and limitations.  ...  Probabilistic programming is a powerful abstraction for statistical machine learning.  ...  Purposes of applying Static Analysis to Probabilistic Programming Static analysis can be applied to probabilistic programs for a variety of reasons.  ... 
arXiv:1909.05076v1 fatcat:4rld375ztvbidm2nbqjpkbqzwy

Conditional independence by typing [article]

Maria I. Gorinova, Andrew D. Gordon, Charles Sutton, Matthijs Vakar
2020 arXiv   pre-print
We present an information flow type system for probabilistic programming that captures conditional independence (CI) relationships, and show that, for a well-typed program in our system, the distribution  ...  A central goal of probabilistic programming languages (PPLs) is to separate modelling from inference. However, this goal is hard to achieve in practice.  ...  In addition, languages like Gen and Turing [Ge et al. 2018] facilitate composable and programmable inference [Mansinghka et al. 2018] , where the user is provided with inference building blocks to implement  ... 
arXiv:2010.11887v1 fatcat:faez77gfp5e5rfdny4inpz4jva
« Previous Showing results 1 — 15 out of 1,222 results