Data driven scientific discovery

Learn more about our research in handling and analysing large and complex data sets to facilitate discovery.

Led by Professor Jarrod Hurley, this program creates methodologies for handling and analysing large and complex data sets to facilitate the data-to-discovery process. This is achieved through research groups, international collaboration and the scientific computing industry.

Our focus is on applications of data-driven computing in various scientific domains, such as biology, physics, astronomy, economics and social sciences. The Swinburne OzSTAR supercomputer is a key piece of infrastructure underpinning this program.


Themes

Novel data science support in scalable hardware and software architectures

Our work in this area supports numerous fields, in particular astronomy. We provide novel data science support in areas such as scalable hardware and software architectures to simulate the growth of structure in the Universe, advanced data management techniques in handling the vast amounts of data generated by all-sky surveys, and advanced machine learning algorithms to sort through these.




Advanced data management and machine learning techniques for all-sky survey data

As astronomy moves ever-closer to the Square Kilometre Array's exascale data era, an increasing number of existing desktop-based workflows will fail.

Instead, astronomers will turn to automated processing using dedicated high-performance computing resources coupled with advanced data archives. Our work focuses on interactive, multi-dimensional analysis of data from observations, simulations, model fits and empirical relationships. Emerging machine learning techniques – such as deep learning – are used to enhance and accelerate the path to discovery.‌


Data visualisation techniques to explore and mine large data sets

Data visualisation is a fundamental knowledge discovery process that is effectively used in many of Swinburne's research projects. It can be an exploratory process, looking for patterns or relationships, or a confirmatory process to support the use of statistical model fitting. Successful data visualisation relies on the use of an appropriate combination of hardware and software.

We work on techniques for large-format two- and three-dimensional (stereoscopic) displays, tiled display walls, virtual reality headsets, and the use of cloud-based virtual desktop infrastructure, with special emphasis on astronomy and health.‌

Associated researchers

Go to Professor Jarrod Hurley page Professor Jarrod Hurley

View Jarrod's full profile

Contact the Data Science Research Institute

There are many ways to engage with us. If your organisation is dealing with a complex problem, then get in touch to discuss how we can work together to provide solutions.

Call +61 3 9214 8180