Exploring the nature of dark energy with modified gravity and machine learning

• Autores: Rubén Arjona Fernández
• Directores de la Tesis: Savvas Nesseris (dir. tes.)
• Lectura: En la Universidad Autónoma de Madrid ( España ) en 2021
• Idioma: inglés
• Número de páginas: 275
• Títulos paralelos:
  • Explorando la naturaleza de la energía oscura con gravedad modificada y aprendizaje automático
• Tribunal Calificador de la Tesis: Antonio López Maroto (presid.), Matteo Martinelli (secret.), Lucas Lombriser (voc.)
• Programa de doctorado: Programa de Doctorado en Física Teórica por la Universidad Autónoma de Madrid
• Materias:
  • Astronomía y astrofísica
    • Cosmología y cosmogonía
  • Física
    • Física teórica
• Enlaces
  • Tesis en acceso abierto en: Biblos-e Archivo
• Resumen

  • This dissertation is focused in exploring the nature of the origin of the accelerated expansion of the Universe at late times from the theoretical point of view through modifications of the theory of gravity and in the analysis of cosmological data in a model independent way through Machine Learning (ML) algorithms, with the goal of testing dark energy (DE) models accurately and probing fundamental properties of gravity.

    In view of the plethora of DE and Modified Gravity (MG) models, the first part of the thesis presents a unified framework to map the MG models, to linear order, to some DE fluid via the effective fluid approach. Then, MG models can be interpreted as DE fluids described by an equation of state, a pressure perturbation and an anisotropic stress. We later show how to implement this approach in the Boltzmann solver CLASS code for f(R) and Horndeski theories respectively obtaining competitive results in a much simpler and less error-prone approach compared to other MG codes like hi_class.

    In the second part of the thesis we use ML algorithms to perform model independent reconstructions of both the background expansion of the Universe but also the perturbations of matter on large scales with methods such as the Genetic Algorithms (GA), which can be best described as a stochastic search approach. The originality of our analysis is that we use a totally agnostic and non-parametric approach based on ML to explore the nature of dark energy and reconstruct its properties in a model independent fashion, which is much broader than traditional statistical inference and model selection.