Empirical evaluation of normalizing flows in Markov chain Monte Carlo

Nabergoj, David (author), Štrumbelj, Erik (author)

Recent advances in MCMC use normalizing flows to precondition target distributions and enable jumps to distant regions. However, there is currently no systematic comparison of different normalizing flow architectures for MCMC. As such, many works choose simple flow architectures that are readily available and do not consider other models. Guidelines for choosing an appropriate architecture would reduce analysis time for practitioners and motivate researchers to take the recommended models as foundations to be improved. We provide the first such guideline by extensively evaluating many normalizing flow architectures on various flow-based MCMC methods and target distributions. When the target density gradient is available, we show that flow-based MCMC outperforms classic MCMC for suitable NF architecture choices with minor hyperparameter tuning. When the gradient is unavailable, flow-based MCMC wins with off-the-shelf architectures. We find contractive residual flows to be the best general-purpose models with relatively low sensitivity to hyperparameter choice. We also provide various insights into normalizing flow behavior within MCMC when varying their hyperparameters, properties of target distributions, and the overall computational budget.

• Language: English
• Keywords: normalizing flow, Markov chain Monte Carlo, comparison, sampling, simulation study
• Work type: Article
• Typology: 1.01 - Original Scientific Article
• Organization: FRI - Faculty of Computer and Information Science
• Publication status: Published
• Publication version: Version of Record
• Year: 2025
• Number of pages: 50 str.
• Numbering: Vol. 114, iss. 12, art. 282
• PID: 20.500.12556/RUL-176286
• UDC: 519.245:004.8
• ISSN on article: 0885-6125
• DOI: 10.1007/s10994-025-06900-3
• COBISS.SI-ID: 258064387
• Publication date in RUL: 26.11.2025

Record is a part of a journal

• Title: Machine learning
• Shortened title: Mach. learn.
• Publisher: Springer Nature
• ISSN: 0885-6125
• COBISS.SI-ID: 2623527

Licences

• License: CC BY 4.0, Creative Commons Attribution 4.0 International
• Link: http://creativecommons.org/licenses/by/4.0/
• Description: This is the standard Creative Commons license that gives others maximum freedom to do what they want with the work as long as they credit the author.

Secondary language

• Language: Slovenian
• Keywords: normalizacijski tok, metode Monte Carlo markovske verige, primerjava, vzorčenje, simulacijska študija

Projects

• Funder: ARIS - Slovenian Research and Innovation Agency
• Project number: P2-0442
• Name: Podatkovne vede in digitalna preobrazba