Self-learning machines hunt for explosions in the universe

The CORTEX consortium of 12 partners from academia, industry and society will make self-learning machines faster, to figure out how massive cosmic explosions work, and to innovate wider applications.

Machine learning has rapidly become an integral part of society, in speech recognition or information retrieval. This is also the case in science, for detecting patterns in nature and the Universe. But the need is growing rapidly for such machines to respond quickly, in the application of self-driving cars and responsive manufacturing for example. On a more fundamental level, self-learning machines help us unveil a dynamical Universe we did not know existed up to recently. Bright explosions appear all over the radio and gravitational-wave sky. Many citizens and scientists are curious to understand where these come from.

“In CORTEX we aim to solve these open problems by bridging fundamental research to society,“ says Dr. Joeri van Leeuwen (ASTRON), the project lead. “We can only reach these ambitious goals if academic, applied, public and industry partners work together.”

Ben van Werkhoven (Netherlands eScience Center) explains the role of the Netherlands eScience center within the project (video in Dutch):

“Within CORTEX, The Netherlands eScience Center investigates how we can create software with the help of AI that can make optimal use of the computing power of modern computers. We then want to apply this technology to implement software with which we can observe explosive events in the universe.”

The Netherlands eScience Center has a central role in CORTEX. We will be extending Kernel Tuner, a tool by Ben van Werkhoven that uses machine learning algorithms to effectively speedup the optimization process of compute-intensive applications, with many new features and capabilities. We will then use this technology to automatically optimize the real-time machine learning pipelines for observing the explosive universe developed within CORTEX.

The 5 million Euro grant from the Nationale Wetenschapsagenda: Onderzoek op Routes door Consortia (NWA-ORC) program thus funds research at partners ASTRON, Nikhef, SURF, Netherlands eScience Center, Universiteit van Amsterdam, Radboud Universiteit Nijmegen, Centrum Wiskunde & Informatica, IBM Nederland B.V., BrainCreators B.V., ABN AMRO N.V., NVIDIA, NOVA, and Stichting ILT; in cooperation with Rijksmuseum, Thermo Fisher B.V., and Leiden University.