Robust Simulation-Based Inference in Cosmology with Bayesian Neural Networks

TL;DR

This paper introduces a Bayesian neural network approach for simulation-based inference in cosmology, addressing generalization issues and biases, and demonstrates its effectiveness with the cosmoSWAG method applied to cosmic microwave background data.

Contribution

It presents the first application of Stochastic Weight Averaging in cosmology, improving the robustness of SBI models against out-of-distribution data.

Findings

Bayesian neural networks reduce biases in SBI.

cosmoSWAG enhances inference reliability.

Application to cosmic microwave background data shows improved performance.

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.