Constraining cosmological parameters from N-body simulations with Variational Bayesian Neural Networks

Abstract

Methods based on Deep Learning have recently been applied on astrophysical parameter recovery thanks to their ability to capture information from complex data. One of these methods is the approximate Bayesian Neural Networks (BNNs) which have demonstrated to yield consistent posterior distribution into the parameter space, helpful for uncertainty quantification. However, as any modern neural networks, they tend to produce overly confident uncertainty estimates and can introduce bias when BNNs are applied to data. In this work, we implement multiplicative normalizing flows (MNFs), a family of approximate posteriors for the parameters of BNNs with the purpose of enhancing the flexibility of the variational posterior distribution, to extract ${\Omega }_m$, $h$, and ${\sigma }_8$ from the QUIJOTE simulations. We have compared this method with respect to the standard BNNs, and the flipout estimator. We found that MNFs combined with BNNs outperform the other models obtaining predictive performance with almost one order of magnitude larger that standard BNNs, ${\sigma }_8$ extracted with high accuracy ($r^2=0.99$), and precise uncertainty estimates. The latter implies that MNFs provide more realistic predictive distribution closer to the true posterior mitigating the bias introduced by the variational approximation and allowing to work with well-calibrated networks.

15 pages, 4 figures, 3 tables, submitted. Comments welcome

References in corpus (6)

• Extracting cosmological parameters from N-body simulations using machine learning techniques
• Constraining the Reionization History using Bayesian Normalizing Flows
• Machine Learning and Cosmology
• Machine-learning cosmology from void properties
• Bayesian Neural Networks: Essentials
• Constraining cosmological parameters from N-body simulations with Bayesian Neural Networks

Cited by in corpus (6)

• Forecasting VIX using Bayesian Deep Learning
• Bayesian deep learning for cosmic volumes with modified gravity
• Accurately Estimating Redshifts from CSST Slitless Spectroscopic Survey using Deep Learning
• Inferring Cosmological Parameters on SDSS via Domain-Generalized Neural Networks and Lightcone Simulations
• Dark from light (DfL): Inferring halo properties from luminous tracers with machine learning trained on cosmological simulations. I. Method, proof of concept & preliminary testing
• Cosmology with Persistent Homology: Parameter Inference via Machine Learning