Cirta is a new machine-learning challenge for high-energy physics on Zindi, the Africa-based data-science challenge platform. Launched this autumn at the International Conference on High Energy and Astroparticle Physics (TIC-HEAP), Constantine, Algeria, Cirta challenges participants to provide machine-learning solutions for identifying particles in LHC experiment data.
Cirta* is the first particle-physics challenge to specifically target computer scientists in Africa, and puts the public TrackML challenge dataset to new use. Created by ATLAS computer scientists Sabrina Amrouche and Dalila Salamani, the Cirta challenge aims to bring new blood into the growing field of machine learning for particle physics.
“Zindi has a strong community of computer scientists based on the continent, and we’re looking forward to reviewing their creative solutions to the challenge,” says Salamani. “Cirta is the first Zindi challenge to focus on physics, and we’re hoping it will widen the community’s focus to new scientific problems.” Amrouche and Salamani themselves studied computer science at the Ecole Nationale Supérieure d'Informatique (ESI) in Algeria, and are now applying their expertise to particle physics as ATLAS PhD students with the University of Geneva.
Through the Cirta challenge, Amrouche and Salamani are also aiming to introduce machine learning to a potential new audience: physics students. The challenge kicked off with a workshop at the TIC-HEAP conference, aimed primarily at Master’s and undergraduate physics students. The workshop encouraged new talents to try their hand at machine learning.
“Despite its rapid growth worldwide, many students have never even heard of machine learning – let alone studied it,” explains Amrouche. “We ended up with an interesting mix of participants at the workshop: some students were comfortable with the software and could discuss the challenge immediately, while others had to be taught from scratch how to use it.”
“But one thing they all had in common was enthusiasm,” adds Amrouche. “Once they saw the power of machine learning and learned how to use it for themselves, they quickly started envisioning potential applications.” Some students have already submitted their solutions to the platform. Others have even planned to make the challenge and machine learning the basis for their master thesis!
Developing machine learning skills can be especially valuable to scientists in developing countries. The software evolution of the past decade has shown that data analysis can be done quickly and efficiently, without the need for massive computing infrastructure. Better software, rather than more hardware, could provide new solutions to present and future problems.
Building on the success of the first workshop, Amrouche and Salamani will be running another machine-learning event in Msila, Algeria in January 2020. “We’re betting on the future,” concludes Salamani. “Through the Cirta challenge, we hope to lay the groundwork to grow the machine learning community in Algeria, and Africa in general. As more people realise its potential - especially for particle physics - we look forward to launching even more complex challenges.”
The Cirta challenge will be open to participants from around the world until 17 February 2020. The top 10 submissions will receive up to 2000 Zindi points. Visit the challenge website to learn more.
* Connecting ancient history to modernity, the Cirta challenge was named after the pre-Roman city now known as Constantine, Algeria.