Swift: Compiled Inference for Probabilistic Programming Languages

06/30/2016
by   Yi Wu, et al.
0

A probabilistic program defines a probability measure over its semantic structures. One common goal of probabilistic programming languages (PPLs) is to compute posterior probabilities for arbitrary models and queries, given observed evidence, using a generic inference engine. Most PPL inference engines---even the compiled ones---incur significant runtime interpretation overhead, especially for contingent and open-universe models. This paper describes Swift, a compiler for the BLOG PPL. Swift-generated code incorporates optimizations that eliminate interpretation overhead, maintain dynamic dependencies efficiently, and handle memory management for possible worlds of varying sizes. Experiments comparing Swift with other PPL engines on a variety of inference problems demonstrate speedups ranging from 12x to 326x.

READ FULL TEXT

page 1

page 2

page 3

page 4

research
07/04/2016

Automatic Generation of Probabilistic Programming from Time Series Data

Probabilistic programming languages represent complex data with intermin...
research
10/02/2018

Automated learning with a probabilistic programming language: Birch

This work offers a broad perspective on probabilistic modeling and infer...
research
02/25/2023

Suspension Analysis and Selective Continuation-Passing Style for Higher-Order Probabilistic Programming Languages

Probabilistic programming languages (PPLs) make encoding and automatical...
research
05/12/2018

π: Towards a Simple Formal Semantic Framework for Compiler Construction

This paper proposes π, a formal semantic framework for compiler construc...
research
10/17/2020

PPL Bench: Evaluation Framework For Probabilistic Programming Languages

We introduce PPL Bench, a new benchmark for evaluating Probabilistic Pro...
research
07/08/2019

Etalumis: Bringing Probabilistic Programming to Scientific Simulators at Scale

Probabilistic programming languages (PPLs) are receiving widespread atte...
research
07/20/2019

Towards Verified Stochastic Variational Inference for Probabilistic Programs

Probabilistic programming is the idea of writing models from statistics ...

Please sign up or login with your details

Forgot password? Click here to reset