Thesis

Simulation inference and generative artificial intelligence for the next era of fundamental physics from observational cosmology

• Call:

  PT-CERN Call 2020/2

• Academic Year:

  2020/2021

• Domain:

  Astroparticle Physics

• Supervisor:

  Andrew Liddle

• Co-Supervisor:

  Antonio da Silva

• Institution:

  FCUL (Universidade de Lisboa)

• Host Institution:

  IA - Instituto de Astrofísica e Ciências do Espaço

• Abstract:

  The overall objective of this project is development, implementation and deployment of novel algorithms for statistical analysis of large cosmological survey datasets, with particular emphasis on machine/deep learning innovations emerging in the literature. These include, but are not limited to; generative models, Bayesian neural networks, intelligent sampling algorithms, and model-independent inference. The application of methods in the emergent literature of artificial intelligence are likely to be amongst the first of their kind in cosmology and physics. The unique power of these machine learning techniques ensures a conclusive and precise search into the future tenets of cosmology. By exploiting the supervisors’ roles in the ongoing Dark Energy Survey and the upcoming Euclid satellite -- co-supervisor António da Silva is the Portuguese lead investigator for Euclid -- we will access state-of-the-art data to target specific science goals, including the nature of primordial perturbations in the Universe and the properties of the neutrino sector. Using innovations brought to the field by artificial intelligence, together with the deep field Euclid observations, the project aims to disseminate advances from understanding early Universe perturbations, exotic cosmologies, theories of gravity and fundamental physics. These investigations, trialled on the Dark Energy Survey, will benefit and shape the understanding of our Universe as informed by the forthcoming Euclid satellite data.