Sat 21 Nov 2020 01:40 - 02:00 at SPLASH-I - F-4A Chair(s): Hidehiko Masuhara
Probabilistic programming languages (PPLs) are an expressive means of representing and reasoning about probabilistic models. The computational challenge of {\em probabilistic inference} remains the primary roadblock for applying PPLs in practice. Inference is fundamentally hard, so there is no one-size-fits all solution. In this work, we target scalable inference for an important class of probabilistic programs: those whose probability distributions are {\em discrete}. Discrete distributions are common in many fields, including text analysis, network verification, artificial intelligence, and graph analysis, but they prove to be challenging for existing PPLs.
We develop a domain-specific probabilistic programming language called Dice that features a new approach to exact discrete probabilistic program inference. Dice exploits program structure in order to factorize inference, enabling us to perform exact inference on probabilistic programs with hundreds of thousands of random variables. Our key technical contribution is a new reduction from discrete probabilistic programs to weighted model counting (WMC). This reduction separates the structure of the distribution from its parameters, enabling logical reasoning tools to exploit that structure for probabilistic inference. We (1) show how to compositionally reduce Dice inference to WMC, (2) prove this compilation correct with respect to a denotational semantics, (3) empirically demonstrate the performance benefits over prior approaches, and (4) analyze the types of structure that allow Dice to scale to large probabilistic programs.
Fri 20 Nov Times are displayed in time zone: Central Time (US & Canada) change
13:00 - 14:20: F-4AOOPSLA at SPLASH-I +12h Chair(s): Ruben MartinsCarnegie Mellon University, Louis MandelIBM Research, USA | |||
13:00 - 13:20 Talk | A Modular Cost Analysis for Probabilistic Programs OOPSLA Link to publication DOI Media Attached | ||
13:20 - 13:40 Talk | Interactive Synthesis of Temporal Specifications from Examples and Natural Language OOPSLA Link to publication DOI Media Attached | ||
13:40 - 14:00 Talk | Scaling Exact Inference for Discrete Probabilistic Programs OOPSLA Steven HoltzenUniversity of California at Los Angeles, Guy Van den BroeckUniversity of California at Los Angeles, Todd MillsteinUniversity of California at Los Angeles Link to publication DOI Pre-print Media Attached | ||
14:00 - 14:20 Talk | Digging for Fold: Synthesis-Aided API Discovery for Haskell OOPSLA Michael B. JamesUniversity of California at San Diego, Zheng GuoUniversity of California, San Diego, Ziteng WangUniversity of California at San Diego, Shivani DoshiUniversity of California at San Diego, Hila PelegUniversity of California at San Diego, Ranjit JhalaUniversity of California at San Diego, Nadia PolikarpovaUniversity of California at San Diego Link to publication DOI Media Attached |