Robust Simulation-Based Inference in Cosmology with Bayesian Neural Networks

Abstract

Simulation-based inference (SBI) is rapidly establishing itself as a standard machine learning technique for analyzing data in cosmological surveys. Despite continual improvements to the quality of density estimation by learned models, applications of such techniques to real data are entirely reliant on the generalization power of neural networks far outside the training distribution, which is mostly unconstrained. Due to the imperfections in scientist-created simulations, and the large computational expense of generating all possible parameter combinations, SBI methods in cosmology are vulnerable to such generalization issues. Here, we discuss the effects of both issues, and show how using a Bayesian neural network framework for training SBI can mitigate biases, and result in more reliable inference outside the training set. We introduce cosmoSWAG, the first application of Stochastic Weight Averaging to cosmology, and apply it to SBI trained for inference on the cosmic microwave background.

5 pages, 3 figures. Preliminary version accepted at the ML4Astro Machine Learning for Astrophysics Workshop at the Thirty-ninth International Conference on Machine Learning (ICML 2022). Final version published at Machine Learning: Science and Technology

References in corpus (4)

• Probabilistic Backpropagation for Scalable Learning of Bayesian Neural Networks
• Uncertainties in Parameters Estimated with Neural Networks: Application to Strong Gravitational Lensing
• Nuisance hardened data compression for fast likelihood-free inference
• Massive data compression for parameter-dependent covariance matrices

Cited by in corpus (16)

• LtU-ILI: An All-in-One Framework for Implicit Inference in Astrophysics and Cosmology
• $\mathrm{S}\mathrm{IM}\mathrm{BIG}$: Mock Challenge for a Forward Modeling Approach to Galaxy Clustering
• KiDS-SBI: Simulation-based inference analysis of KiDS-1000 cosmic shear
• GLASS: Generator for Large Scale Structure
• EFTofLSS meets simulation-based inference: ${\sigma }_8$ from biased tracers
• Fast and robust Bayesian Inference using Gaussian Processes with GPry
• Stochastic Modelling of Star Formation Histories III. Constraints from Physically-Motivated Gaussian Processes
• Jensen-Shannon Divergence Based Novel Loss Functions for Bayesian Neural Networks
• Cosmology with Persistent Homology: a Fisher Forecast
• Scalable inference with Autoregressive Neural Ratio Estimation
• Isolated pulsar population synthesis with simulation-based inference
• Generalised likelihood profiles for models with intractable likelihoods
• Simulation-based inference of single-molecule force spectroscopy
• Neural Posterior Estimation for Cataloging Astronomical Images with Spatially Varying Backgrounds and Point Spread Functions
• Towards characterizing dark matter subhalo perturbations in stellar streams with graph neural networks
• Quantifying Weighted Morphological Content of Large-Scale Structures via Simulation-Based Inference