Bayesian modeling of strong gravitational lenses observed with Euclid and JWST - Code Optimisation and Parallelisation

Project Description

The present project aims to port a high-performance code for the study and modeling of strong gravitational lenses on highly parallel architectures and GPUs. The code is entirely based on Bayesian principles and is implemented in Julia, a programming language that offers many advantages in the scientific field and that lends itself very well to the development of parallel code. The realization of the project will require a significant investment of time, largely related to the implementation of the code and porting to parallel architectures, and the purchase of a server with a scientific GPU card dedicated to data analysis. Upon its completion, the proposed project will be able to bridge the gap that currently exists between the complexity of optical data produced by new generation space telescopes, such as Euclid and JWST, and the current capabilities of gravitational lens modeling. In this way, it will finally be possible to analyze complex gravitational lens systems such as galaxy clusters based on the use of extended images and giant arcs with tens or hundreds of multiple image families, and it will be possible to verify through hypothesis testing based on Bayesian techniques the presence of substructures in the dark matter halos of the lens.

Goal

The main goal of the project is the production of a highly parallel code capable of performing, in a short time, the Bayesian inference of gravitational lens systems that include hundreds of free parameters. The project starts from an already available code, Gravity, developed by Prof. M. Lombardi., which already allows Bayesian inference on gravitational lens systems. The code contains some innovative elements, which make it already competitive with others available in the literature.

The project has also the goal of publishing a paper on the work done.

Who we are looking for

• We are looking for a master student (close to finishing the master or with the master recently obtained) in a scientific field, as physics, chemistry, engineering or computer science for an internship. 

• We expect experience in programming in Python. In particular experience in Julia is a real plus. 

• Duration March until September, for a 50-80% position.

Who we are

TOELT LLC is a company that focus on fundamental research in various scientific applications of machine learning ( www.toelt.ai ). We work in various fields, from astrophysics, Spectroscopy, food technology, bio-engineering, medicine, deep learning theory, statistics and deep learning theory. We publish regularly in peer reviewed journals, and are active internationally.