Variational inference as an alternative to MCMC for parameter estimation and model selection

Gunapati, Geetakrishnasai and Jain, Anirudh and Srijith, P. K. and Desai, Shantanu (2022) Variational inference as an alternative to MCMC for parameter estimation and model selection. Publications of the Astronomical Society of Australia, 39. ISSN 1323-3580

Full text not available from this repository. (Request a copy)

Abstract

Most applications of Bayesian Inference for parameter estimation and model selection in astrophysics involve the use of Monte Carlo techniques such as Markov Chain Monte Carlo (MCMC) and nested sampling. However, these techniques are time-consuming and their convergence to the posterior could be difficult to determine. In this study, we advocate variational inference as an alternative to solve the above problems, and demonstrate its usefulness for parameter estimation and model selection in astrophysics. Variational inference converts the inference problem into an optimisation problem by approximating the posterior from a known family of distributions and using Kullback-Leibler divergence to characterise the difference. It takes advantage of fast optimisation techniques, which make it ideal to deal with large datasets and makes it trivial to parallelise on a multicore platform. We also derive a new approximate evidence estimation based on variational posterior, and importance sampling technique called posterior-weighted importance sampling for the calculation of evidence, which is useful to perform Bayesian model selection. As a proof of principle, we apply variational inference to five different problems in astrophysics, where Monte Carlo techniques were previously used. These include assessment of significance of annual modulation in the COSINE-100 dark matter experiment, measuring exoplanet orbital parameters from radial velocity data, tests of periodicities in measurements of Newton's constant G, assessing the significance of a turnover in the spectral lag data of GRB 160625B, and estimating the mass of a galaxy cluster using weak gravitational lensing. We find that variational inference is much faster than MCMC and nested sampling techniques for most of these problems while providing competitive results. All our analysis codes have been made publicly available. ©

[error in script]

IITH Creators:

IITH Creators	ORCiD
Srijith, P K	https://orcid.org/0000-0002-2820-0835
Desai, Shantanu	http://orcid.org/0000-0002-0466-3288

Item Type:

Article

Additional Information:

Geetakrishnasai Gunapati was supported by DST-ICPS grant. Anirudh Jain was supported by the Microsoft summer internship program at IIT Hyderabad. We would like to thank Daniel Gruen, Soumya Mohanty, Sanjib Sharma, and Jochen Weller for useful correspondence and making available to us some of the datasets used in this work. a

Uncontrolled Keywords:

Astronomy data analysis; Bayesian model comparison

Subjects:

Computer science
Physics

Divisions:

Department of Computer Science & Engineering
Department of Physics

Depositing User:

. LibTrainee 2021

Date Deposited:

14 Jul 2022 11:27

Last Modified:

14 Jul 2022 11:27