Likelihood-free Model Selection in Cosmic Reionization with Three-dimensional Tomographic 21 cm Lightcone Images

TL;DR

This paper applies likelihood-free Bayesian model selection using CNNs and simulation-based methods to distinguish reionisation scenarios from 21cm lightcone images, demonstrating promising results but highlighting challenges in accurately modeling outside-in reionisation.

Contribution

It introduces a likelihood-free model selection approach with CNNs and Bayesian evidence calculation for cosmic reionization analysis, emphasizing the need for flexible summarization methods.

Findings

Successfully distinguishes models used for mock data

Struggles with accurate posteriors for outside-in models

Highlights importance of flexible model summarization

Abstract

We explore likelihood-free (aka simulation-based) Bayesian model selection to quantify model comparison analyses of reionisation scenarios. We iteratively train the 3D Convolutional Neural Network (CNN) on four toy EoR models based on 21cmFAST simulations with contrasting morphology to obtain summaries of the 21 cm lightcone. Within the pyDelfi framework, we replaced the Emcee sampler with MultiNest to integrate learnt posteriors and produce the Bayesian Evidence. We comfortably distinguish the model used to produce the mock data set in all cases. However, we struggle to produce accurate posterior distributions for outside-in reionisation models. After a variety of cross-checks and alternate analyses we discuss the flexibility of summarising models that differ from precisely the intended network training conditions as this should be more widely scrutinised before CNN can reliably analyse observed data.