©Claudia Marcelloni - CERN
Gilles Louppe, a lecturer in Artificial Intelligence within the Montefiore Institute (Electrical Engineering and Computer Science dpt) , is involved in the development of faster and more reliable analytical methods to detect the most furtive physical phenomena.
W
ith physicists from New York University, Gilles Louppe, now a lecturer in Artificial Intelligence at ULiège, has participated in the development of new techniques based on Artificial Intelligence and automatic learning to considerably improve the analysis of data generated within the Large Hadron Collider (LHC), the most powerful particle accelerator in the world located near Geneva straddling the Swiss and French borders.
The study of the "new physics" is at stake at the LHC, a CERN laboratory. Every second in the heart of the underground ring, millions of proton collisions occur, generating an incredible amount of data. Detecting and measuring stealth processes of interest to physics is almost like looking for a needle in a haystack...
The LHC is exploring a new frontier in high-energy physics. It could reveal the origin of the mass of fundamental particles, the source of the dark matter that fills the universe, and even other dimensions of space. It is there that in 2012, based on the data collected, physicists were able to confirm the existence of the Higgs boson, a subatomic particle that plays a key role in our understanding of the universe and whose existence had been described for several years by the Belgian professor François Englert (ULB), which earned the Nobel Prize in Physics with Peter Higgs.
The new analytical methods recently presented in the prestigious journal Physical Review D are inspired by Artificial Intelligence. They have already been implemented by American researchers to successfully attest to the existence of the Higgs boson. Today they are promising to allow the discovery of many other phenomena and physical particles unsuspected until now.
The combination of data science, computer science and physics provides here new artificial intelligence tools on which researchers will rely to make new discoveries and revolutionize fundamental knowledge in physics.
It was during his post-doctoral stay at CERN that Gilles Louppe began this research with his colleagues at New York University, which he is now pursuing at the University of Liège. He is co-author of the papers in Physical Review D.
Gilles Louppe explains: "Our methods have been developed primarily for high-energy physics, but they are general enough to be applied to many other problems. In many scientific fields, computer simulations indeed often provide the most accurate description of a complicated phenomenon. However, these simulations are difficult to use directly in the context of a scientific analysis. In this work, we propose an effective algorithm based on deep learning, allowing these high-precision simulations to be used to carry out data analysis."
" To make things more concrete, for instance, it is very easy to write a computer program
that would simulate a pool game where the balls bounce against each other and against the edges of the table. However, it is much more complicated to look at the final arrangement of the balls and infer the angle at which the cue ball was struck. Similarly, while we often imagine blackboards filled with equations, modern physics is built on computer simulators. These simulations can be very accurate, but they do not immediately enable the analysis of experimental observations."
Scientific reference
Johann Brehmer, Kyle Cranmer, Gilles Louppe, and Juan Pavez, A guide to constraining effective field theories with machine learning, Phys. Rev. D 98, 052004 – September 2018.
Contact
Gilles LOUPPE, Montefiore Institute of Electrical Engineering and Computer Science
