Enhancing Cosmological Model Selection with Interpretable Machine Learning

TL;DR

This paper introduces neural networks combined with interpretability methods to effectively differentiate between cosmological models using large-scale structure data, achieving high accuracy and aiding in probing deviations from general relativity.

Contribution

The study demonstrates the successful application of neural networks with interpretability techniques to distinguish cosmological models, enhancing analysis of survey data for fundamental physics.

Findings

Neural networks achieved approximately 97% accuracy in model classification.

LIME interpretability identified key features influencing model decisions.

The approach improves the potential of survey data to test deviations from general relativity.

Abstract

We propose a novel approach using neural networks (NNs) to differentiate between cosmological models, and implemented LIME as an interpretability approach to identify the key features influencing our model's decisions. We show the potential of NNs to enhance the extraction of meaningful information from cosmological large-scale structure data, based on current galaxy-clustering survey specifications, for the cosmological constant and cold dark matter ($\Lambda$CDM) model and the Hu-Sawicki $f(R)$ model. We find that the NN can successfully distinguish between $\Lambda$CDM and the $f(R)$ models, by predicting the correct model with approximately $97\%$ overall accuracy, thus demonstrating that NNs can maximize the potential of current and next generation surveys to probe for deviations from general relativity.