Principal Component and Regression Analyses on Nuclear Equation of State Parameters Constrained by Neutron Star Observations

Document Type


Degree Name

Master of Science (MS)


Physics and Astronomy

Date of Award

Summer 2023


Machine learning algorithms have grown in popularity in the physical sciences in accordance with increasing computational power. This is especially true in nuclear and higher energy physics as more experiments produce large data sets which need to be studied carefully and efficiently. One particularly interesting research question in nuclear physics is how to model and describe the nuclear equation of state (EoS), especially at saturation density. This is an especially ripe area in nuclear physics because there are terrestrial experiments and astronomical observations that can provide direct constraints on the EoS. In literature these constraints are provided by different Monte Carlo Markov Chain (MCMC) processes to infer parameterized EoS which can then be used to restrict theoretical and phenomenological predictions. The MCMC process outputs a joint posterior probability distribution of the input parameters which were constrained by comparing the theoretical predictions with a given set of observations. Then, generally, the probability distribution is marginalized for each model parameter and often simple Pearson correlation scores are provided between the parameters. In this work we discuss some basic machine learning algorithms and the EOS parametrizations and use these as a guide to explore the potential use of a well established machine learning algorithms, principal component analysis, and various regression techniques as an extension to better understand the correlations between EOS parameters constrained by neutron star data. We find a number of statistically viable models and conclude the importance of each EOS parameter in determining the radius of a 1.4 solar mass neutron star. In order we determined these to be Ksym, L, Jsym, K0, J0, and Esym.


Bao-An Li

Subject Categories

Astrophysics and Astronomy | Computer Sciences | Physical Sciences and Mathematics