Abstract
To extract signals from the Cosmic Microwave Background (CMB) and the Cosmic Infrared Background (CIB) properly, it is important that we have proper models of the foreground emission. One way to improve our models is to identify degeneracies in the parameters of the model. However, identifying degeneracies can be slow, depending on how many parameters we need to explore. The Python module, Zodipy, has models of the Zodiacal light using over fifty parameters, which means that we need a method to explore a parameter space of over fifty dimensions efficiently. In this thesis, we implement and utilize the Hamiltonian Monte Carlo(HMC) method using automatic differentiation through the JAX module, and the Zodipy module to explore the parameter space in the Zodiacal model by Kelsall et al. (1998), which utilizes the DIRBE data set. The model consists of six components: Three Dust Bands, a Circumsolar Ring, the Earth-trailing feature, and a diffuse cloud. We mainly applied the HMC method and explored the parameter space in the diffuse cloud. To explore the parameter space, we applied arbitrary noise to the Zodiacal emission from the DIRBE model to recreate the situation where our emissions are affected by noise during observation. We explored the parameter space in the diffuse cloud and observed that we found some degeneracies between the parameters. From this we can conclude that the parameters affected are both well constrained/highly correlated depending on the parameter it is compared with, in addition to verifying that JAX and Zodipy can be implemented together with caveats.
Supervisors:
Researcher Duncan Watts, Institute of Theoretical Astrophysics, UiO
Professor Ingunn Kathrine Wehus, Institute of Theoretical Astrophysics, UiO
Intern. assessor: Associate Professor Tiago Pereira, RoCS, Institute of Theoretical Astrophysics, UiO
Extern. assessor: Data scientist Trygve Svalheim, Menon Economics