QUB logo

Astrophysics Research Centre

School of Mathematics and Physics


Machine learning and computational techniques in the era of massive data rich surveys

Background

Astronomical science is an increasingly data driven exercise in which we curate, calibrate and analyse imaging data sets of peta-byte scale. This calls for new image recognition techniques that can autonomously classify sources on astronomical images and alert scientists to the existence of new physical phenomena.

Within ARC we have leading roles in several major sky surveys. The ATLAS project is moving to an all-sky (northern and southern hemispheres) survey and will be the first ever survey to map both hemispheres multiple times every single night to sensitive flux levels. We analyse the data rapidly at Queen's while the telescopes are observing in Hawaii (and soon to add Chile and South Africa). We have been partners in the Pan-STARRS sky surveys for many years which is a world leading survey telescope in Hawaii and we partners in the Large Synoptic Survey Telescope which is being constructed in Chile (Prof Smartt is the UK project scientist).

We rapidly search through these images, scanning for new explosive events in the Universe and weed out the false positives and classify the real sources. We use a range of boosted decision trees and machine learning techniques. We pass these discoveries to other large facilities to follow-up. We are also partners in the massive spectroscopic survey 4MOST, and a key area for discovery is combining the imaging with triggering this facility. Another aspect is the management and data mining billion row databases.

The project

The project will work on developing our existing machine learning techniques for the ATLAS, Pan-STARRS and LSST projects and eventually link to the 4MOST spectroscopic survey. This is a multi-disciplinary project that aims at combining state of the art machine learning, database technology and image recognition tools with the latest in astronomical discovery.

We are searching for computationally talented students who are interested in coding and developing our algorithms. Students with undergraduate degrees in maths, physics, computer science or engineering are encouraged to apply.

More info

Supervisor: Prof Stephen Smartt

public/phds2019/2019_ml_smartt.txt · Last modified: 2018/12/20 12:39 by Stephen Smartt


Privacy & Cookies | Accessibility statement
Back to Top Sitemap News