Self-learning machines for better understanding of the universe
Bright explosions appear all over the radio and gravitational-wave sky. This dynamic side of the universe which has just been discovered, can be mapped by self-learning machines. The National Science Agenda granted 5 million euro’s to CORTEX, the Center for Optimal, Real-Time Machine Studies of the Explosive Universe. The Leiden Institute of Advanced Computer Science will conduct part of the fundamental research in CORTEX.
'Self-learning systems have shown us a glimpse of true machine intelligence in games such as Go, now we will apply this technology to exploring the universe,' says Aske Plaat, Professor in Data science. He coordinates the contribution of Leiden to CORTEX in collaboration with Huub Rötgering of the Observatory of Leiden. LIACS will work on the self-learning software to speed up the processing of large amounts of diverse astronomical data, Plaat explains. Combining fundamental knowledge of self-learning systems with knowledge of astronomical data makes this possible.
More than astronomical purposes
CORTEX hopes to develop faster self-learning machines by doing more fundamental research on artificial intelligence. Sarah Caudill of CORTEX-participant Nikhef wants to use these self-learning machines to investigate black holes and neutron stars. ‘The gravitational-waves that are formed when black holes and neutron stars melt together, peaks only for a few seconds. Maybe once a week’, she says. ‘Faster artificial intelligence can help to recognize these explosions before they go off.’
Faster self-learning machines can be used for more than astronomical purposes. The technology can contribute to self-driving cars or responsive manufacturing in the future.
Unique collaboration
Universities and industry collaborate on a close level in CORTEX. The centre finances research in twelve institutions, constisting of universities, applied, public and commercial partners.
Partners
ASTRON, Nikhef, Netherlands eScience Center, University of Amsterdam, Radboud University, Centrum Wiskunde & Informatica, IBM Nederland B.V., BrianCreators B.V., ABN AMRO N.V., NVIDIA, NOVA en Stichting ILT; in collaboration with Rijksmuseum, Thermo Fisher Scientific en Leiden University.