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

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 work, 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 optimization problem by approximating the posterior from a known family of distributions and using Kullback-Leibler divergence to characterize the difference. It takes advantage of fast optimization techniques, which make it ideal to deal with large datasets and makes it trivial to parallelize 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 (PWISE), 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.

arXiv comment17 pages, 3 figures. Added one more example and introduced and applied a new metric for calculating Bayesian evidence in Variational Inference. Accepted for publication in PASA

References in corpus (33)

• emcee: The MCMC Hammer
• Seven-Year Wilkinson Microwave Anisotropy Probe (WMAP) Observations: Cosmological Interpretation
• Variational Inference: A Review for Statisticians
• MultiNest: an efficient and robust Bayesian inference tool for cosmology and particle physics
• dynesty: A Dynamic Nested Sampling Package for Estimating Bayesian Posteriors and Evidences
• Importance Nested Sampling and the MultiNest Algorithm
• Adaptive Refinement Tree - a new high-resolution N-body code for cosmological simulations
• Data analysis recipes: Using Markov Chain Monte Carlo
• Dynamic temperature selection for parallel-tempering in Markov chain Monte Carlo simulations
• On the Accuracy of Weak Lensing Cluster Mass Reconstructions
• Markov Chain Monte Carlo Methods for Bayesian Data Analysis in Astronomy
• Bayesian parameter estimation using conditional variational autoencoders for gravitational-wave astronomy
• Masses of galaxy clusters from gravitational lensing
• Variational inference for large-scale models of discrete choice
• Galaxy Zoo: Probabilistic Morphology through Bayesian CNNs and Active Learning
• Uncertainties in Parameters Estimated with Neural Networks: Application to Strong Gravitational Lensing
• Measurements of Newton's gravitational constant and the length of day
• Bayesian Calibrated Significance Levels Applied to the Spectral Tilt and Hemispherical Asymmetry
• A New Test of Lorentz Invariance Violation: the Spectral Lag Transition of GRB 160625B
• ExoFit: Orbital Parameters of Extra-solar Planets from Radial Velocities
• Parameters Estimation for the Cosmic Microwave Background with Bayesian Neural Networks
• AstroVaDEr: Astronomical Variational Deep Embedder for Unsupervised Morphological Classification of Galaxies and Synthetic Image Generation
• Detection of Gravitational Waves Using Bayesian Neural Networks
• Exoplanet Validation with Machine Learning: 50 new validated Kepler planets
• Statistical Significance of spectral lag transition in GRB 160625B
• The stepping-stone sampling algorithm for calculating the evidence of gravitational wave models
• Constraining the Reionization History using Bayesian Normalizing Flows
• Tracing Halpha Fibrils through Bayesian Deep Learning
• Frequentist model comparison tests of sinusoidal variations in measurements of Newton's gravitational constant
• An independent assessment of significance of annual modulation in COSINE-100 data
• Structured Variational Inference for Simulating Populations of Radio Galaxies
• Comment on "Measurements of Newton's gravitational constant and the length of day"
• Stochastic Gradient Annealed Importance Sampling for Efficient Online Marginal Likelihood Estimation

Cited by in corpus (12)

• A novel framework for semi-Bayesian radial velocities through template matching
• Solving the $H_0$ tension in $f(T)$ Gravity through Bayesian Machine Learning
• Astrophysical and Cosmological Searches for Lorentz Invariance Violation
• Constraining cosmological parameters from N-body simulations with Variational Bayesian Neural Networks
• Field-level simulation-based inference with galaxy catalogs: the impact of systematic effects
• Bayesian deep learning for cosmic volumes with modified gravity
• Bayesian evidence for spectral lag transition due to Lorentz Invariance Violation for 32 Fermi/GBM Gamma-ray Bursts
• Spectral evolution responsible for the transition from positive lags to negative lags in Gamma-ray Bursts
• jaxspec : a fast and robust Python library for X-ray spectral fitting
• Constraint on Lorentz Invariance Violation for spectral lag transition in GRB 160625B using profile likelihood
• High energy time lags of Gamma Ray Bursts
• ProBoost: a Boosting Method for Probabilistic Classifiers