Constraints on Cubic and $f(P)$ Gravity from the Cosmic Chronometers, BAO & CMB datasets : Use of Machine Learning Algorithms

TL;DR

This paper constrains Einsteinian cubic and $f(P)$ gravity models using cosmic chronometers, BAO, and CMB data, employing machine learning algorithms like linear regression and neural networks to estimate the Hubble parameter and compare with observations.

Contribution

It introduces a novel combination of observational data analysis with machine learning techniques to constrain and estimate parameters in advanced gravity models.

Findings

Theoretical models fit observational data well.

Machine learning models accurately estimate $H(z)$.

Parameter bounds are effectively derived using MCMC.

Abstract

In this work, we perform observational data analysis on Einsteinian cubic gravity and $f(P)$ gravity to constrain the parameter space of the theories. We use the 30-point $z-H(z)$ cosmic chronometer data as the observational tool for our analysis along with the BAO and the CMB peak parameters. The $\chi^2$ statistic is used for the fitting analysis and it is minimized to obtain the best fit values for the free model parameters. We have used the Markov chain Monte Carlo algorithm to obtain bounds for the free parameters. To achieve this we used the publicly available CosmoMC code to put parameter bounds and subsequently generate contour plots for them with different confidence intervals. Besides finding the Hubble parameter $H$ in terms of the redshift $z$ theoretically from our gravity models, we have exercised correlation coefficients and two machine learning models, namely the linear regression (LR) and artificial neural network (ANN), for the estimation of $H(z)$. For this purpose, we have developed a Python package for finding the parameter space and performing the subsequent statistical analysis and prediction analysis using machine learning. We compared both our theoretical and estimated values of $H(z)$ with the observations. It is seen that our theoretical and estimated models from machine learning performed significantly well when compared with the observations.