A factor graph approach to automated design of Bayesian signal processing algorithms

Marco Cox (Corresponding author), Thijs van de Laar, Bert de Vries

Research output: Contribution to journalArticleAcademicpeer-review

2 Citations (Scopus)

Abstract

The benefits of automating design cycles for Bayesian inference-based algorithms are becoming increasingly recognized by the machine learning community. As a result, interest in probabilistic programming frameworks has much increased over the past few years. This paper explores a specific probabilistic programming paradigm, namely message passing in Forney-style factor graphs (FFGs), in the context of automated design of efficient Bayesian signal processing algorithms. To this end, we developed “ForneyLab”2 as a Julia toolbox for message passing-based inference in FFGs. We show by example how ForneyLab enables automatic derivation of Bayesian signal processing algorithms, including algorithms for parameter estimation and model comparison. Crucially, due to the modular makeup of the FFG framework, both the model specification and inference methods are readily extensible in ForneyLab. In order to test this framework, we compared variational message passing as implemented by ForneyLab with automatic differentiation variational inference (ADVI) and Monte Carlo methods as implemented by state-of-the-art tools “Edward” and “Stan”. In terms of performance, extensibility and stability issues, ForneyLab appears to enjoy an edge relative to its competitors for automated inference in state-space models.

LanguageEnglish
Pages185-204
Number of pages20
JournalInternational Journal of Approximate Reasoning
Volume104
DOIs
StatePublished - 1 Jan 2019

Fingerprint

Factor Graph
Signal Processing
Signal processing
Message passing
Message Passing
Probabilistic Programming
Differentiation (calculus)
Automatic Differentiation
Model Comparison
Model Specification
State-space Model
Bayesian inference
Parameter estimation
Monte Carlo method
Learning systems
Parameter Estimation
Machine Learning
Monte Carlo methods
Paradigm
Specifications

Bibliographical note

Accepted for publication in the International Journal of Approximate Reasoning

Keywords

  • Bayesian inference
  • Factor graphs
  • Julia
  • Message passing
  • Probabilistic programming

Cite this

@article{29135ae69bdb4c78bdfc9c24e98700fc,
title = "A factor graph approach to automated design of Bayesian signal processing algorithms",
abstract = "The benefits of automating design cycles for Bayesian inference-based algorithms are becoming increasingly recognized by the machine learning community. As a result, interest in probabilistic programming frameworks has much increased over the past few years. This paper explores a specific probabilistic programming paradigm, namely message passing in Forney-style factor graphs (FFGs), in the context of automated design of efficient Bayesian signal processing algorithms. To this end, we developed “ForneyLab”2 as a Julia toolbox for message passing-based inference in FFGs. We show by example how ForneyLab enables automatic derivation of Bayesian signal processing algorithms, including algorithms for parameter estimation and model comparison. Crucially, due to the modular makeup of the FFG framework, both the model specification and inference methods are readily extensible in ForneyLab. In order to test this framework, we compared variational message passing as implemented by ForneyLab with automatic differentiation variational inference (ADVI) and Monte Carlo methods as implemented by state-of-the-art tools “Edward” and “Stan”. In terms of performance, extensibility and stability issues, ForneyLab appears to enjoy an edge relative to its competitors for automated inference in state-space models.",
keywords = "Bayesian inference, Factor graphs, Julia, Message passing, Probabilistic programming",
author = "Marco Cox and {van de Laar}, Thijs and {de Vries}, Bert",
note = "Accepted for publication in the International Journal of Approximate Reasoning",
year = "2019",
month = "1",
day = "1",
doi = "10.1016/j.ijar.2018.11.002",
language = "English",
volume = "104",
pages = "185--204",
journal = "International Journal of Approximate Reasoning",
issn = "0888-613X",
publisher = "Elsevier",

}

TY - JOUR

T1 - A factor graph approach to automated design of Bayesian signal processing algorithms

AU - Cox,Marco

AU - van de Laar,Thijs

AU - de Vries,Bert

N1 - Accepted for publication in the International Journal of Approximate Reasoning

PY - 2019/1/1

Y1 - 2019/1/1

N2 - The benefits of automating design cycles for Bayesian inference-based algorithms are becoming increasingly recognized by the machine learning community. As a result, interest in probabilistic programming frameworks has much increased over the past few years. This paper explores a specific probabilistic programming paradigm, namely message passing in Forney-style factor graphs (FFGs), in the context of automated design of efficient Bayesian signal processing algorithms. To this end, we developed “ForneyLab”2 as a Julia toolbox for message passing-based inference in FFGs. We show by example how ForneyLab enables automatic derivation of Bayesian signal processing algorithms, including algorithms for parameter estimation and model comparison. Crucially, due to the modular makeup of the FFG framework, both the model specification and inference methods are readily extensible in ForneyLab. In order to test this framework, we compared variational message passing as implemented by ForneyLab with automatic differentiation variational inference (ADVI) and Monte Carlo methods as implemented by state-of-the-art tools “Edward” and “Stan”. In terms of performance, extensibility and stability issues, ForneyLab appears to enjoy an edge relative to its competitors for automated inference in state-space models.

AB - The benefits of automating design cycles for Bayesian inference-based algorithms are becoming increasingly recognized by the machine learning community. As a result, interest in probabilistic programming frameworks has much increased over the past few years. This paper explores a specific probabilistic programming paradigm, namely message passing in Forney-style factor graphs (FFGs), in the context of automated design of efficient Bayesian signal processing algorithms. To this end, we developed “ForneyLab”2 as a Julia toolbox for message passing-based inference in FFGs. We show by example how ForneyLab enables automatic derivation of Bayesian signal processing algorithms, including algorithms for parameter estimation and model comparison. Crucially, due to the modular makeup of the FFG framework, both the model specification and inference methods are readily extensible in ForneyLab. In order to test this framework, we compared variational message passing as implemented by ForneyLab with automatic differentiation variational inference (ADVI) and Monte Carlo methods as implemented by state-of-the-art tools “Edward” and “Stan”. In terms of performance, extensibility and stability issues, ForneyLab appears to enjoy an edge relative to its competitors for automated inference in state-space models.

KW - Bayesian inference

KW - Factor graphs

KW - Julia

KW - Message passing

KW - Probabilistic programming

UR - http://www.scopus.com/inward/record.url?scp=85056602296&partnerID=8YFLogxK

U2 - 10.1016/j.ijar.2018.11.002

DO - 10.1016/j.ijar.2018.11.002

M3 - Article

VL - 104

SP - 185

EP - 204

JO - International Journal of Approximate Reasoning

T2 - International Journal of Approximate Reasoning

JF - International Journal of Approximate Reasoning

SN - 0888-613X

ER -