Cosmic Inflation and Genetic Algorithms

Abel, Steve A.; Constantin, Andrei; Harvey, Thomas R.; Lukas, Andre

Abstract

Large classes of standard single-field slow-roll inflationary models consistentwith the required number of e-folds, the current bounds on the spectral indexof scalar perturbations, the tensor-to-scalar ratio, and the scale of inflation canbe efficiently constructed using genetic algorithms. The setup is modular andcan be easily adapted to include further phenomenological constraints. Asemi-comprehensive search for sextic polynomial potentials results in∼(300,000) viable models for inflation. The analysis of this dataset revealsa preference for models with a tensor-to-scalar ratio in the range0.0001≤r≤0.0004. We also consider potentials that involve cosine andexponential terms. In the last part we explore more complex methods ofsearch relying on reinforcement learning and genetic programming. Whilereinforcement learning proves more difficult to use in this context, the geneticprogramming approach has the potential to uncover a multitude of viableinflationary models with new functional forms.

Citation

Abel, S. A., Constantin, A., Harvey, T. R., & Lukas, A. (2023). Cosmic Inflation and Genetic Algorithms. Fortschritte der Physik, 71(1), https://doi.org/10.1002/prop.202200161

• Journal Article Type: Article
• Online Publication Date: Oct 29, 2022
• Publication Date: 2023
• Deposit Date: Jan 16, 2023
• Publicly Available Date: Jan 16, 2023
• Journal: Fortschritte der Physik
• Print ISSN: 0015-8208
• Electronic ISSN: 1521-3978
• Publisher: Wiley-VCH Verlag
• Peer Reviewed: Peer Reviewed
• Volume: 71
• Issue: 1
• DOI: https://doi.org/10.1002/prop.202200161
• Public URL: https://durham-repository.worktribe.com/output/1181701