Data mining and machine learning for natural sciences project

NICTA, Australia’s Information and Communications Technology (ICT) Research Centre of Excellence, is leading a multimillion-dollar project that will use big data and machine learning to deliver new insights into the natural sciences.

The $12 million Big Data Knowledge Discovery project is a three-year research endeavour to advance fundamental mathematics and statistics to provide a framework, methodologies and tools for data-enabled scientific insight and discovery. It is supported by $4 million from the Science and Industry Endowment Fund (SIEF) and $8 million from NICTA’s research collaborators - SIRCA, Macquarie University and The University of Sydney - over the life of the project.

Chief Scientist for Australia Professor Ian Chubb said: “This project positions Australia to one day be home to a new generation of big data analytics tools that could be used by all manner of scientists around the world to advance knowledge discovery.”

The project brings together some of the best computer and data scientists in the world from NICTA (machine learning and analytics) with software engineers from SIRCA (software and big data), along with three of Australia’s most distinguished natural scientists in physics, plant science and geosciences, to tackle grand scientific challenges in completely new ways:

• Terrestrial ecology - led by Professor Mark Westoby (Macquarie University)
• Physics and mathematics of complex laser systems - led by Professor Deb Kane (Macquarie University)
• Geosciences, Earth dynamics and tectonics - led by Professor Dietmar Muller (University of Sydney)

The multidisciplinary team will use data science to determine which ecological interactions are most important in producing the world we see around us, potentially opening a window on some of the mysteries of biodiversity and showing how ecosystems will be affected by climate change and other factors.

Their work in complex laser systems will help improve the security of optical fibre communication systems.

The project will also combine publicly available geological data from Geosciences Australia with SIRCA technology (used to predict stock market movements) to picture what Australia was like 1.5 billion years ago and how its rich metal deposits were formed - allowing for their retrieval.

Professor Muller said investment in high-risk mining exploration in Australia has recently “plummeted”, so “any information that can improve the accuracy of such ‘greenfield’ exploration would obviously be highly valued.”

“Part of the exciting potential of this project is that by applying algorithms used to predict stock market movements to the detailed data sets we have on Australia’s geology, we hope to be able to uncover regions which have the highest likelihood of having exploitable mineral deposits buried hundreds of metres under weathered surface rocks.”

The research will help discover sophisticated data analysis processes that can reduce the amount of raw data needed to conduct successful experiments, potentially hastening the rate of scientific progress.

Speaking at the launch of the project, NICTA CEO Professor Hugh Durrant-Whyte said: “This project will explore a new and powerful paradigm for data-intensive science… ranging from medical and health sciences to physics.”

“Data and the right software can be very powerful catalysts for both breakthrough research and innovation,” said SIRCA CEO Dr Michael Briers.

“Motivated by the grand challenges of big data, this initiative brings together world-leading experts in finance and geoscience - possibly for the first time. SIRCA is very excited by this opportunity for two such diverse disciplines to work collaboratively towards a common goal.”