CA24153 - Edge Deep Learning for Particle Physics (EPIGRAPHY)

Deep Learning applications are becoming ubiquitous in science. Particle physicists in particular are relying more and more on neural networks for crucial tasks in their data processing. Often, they are confronted with unique technological challenges, such as the need to operate deep neural networks in real time and in extreme computing environments, such as the event filtering systems of the experiments at the Large Hadron Collider. Low latency and reduced computing resources imply strong constraints for the algorithms operated on-edge, i.e., as close to the detector as possible. To take advantage of the power of deep learning under these conditions, one needs to develop compression techniques and make the networks efficient without losing their expressivity. This COST action aims at gathering researchers across Europe interested in this challenging problem, to advance AI-powered on-edge inference for next-generation particle physics experiments like the High Luminosity LHC and future colliders. Through the technological advancements envisioned, many other fields and sectors relying on fast AI-based decision making and on-edge computation such as smartphones, automotive, portable medical devices, drones, or satellites may too profit from our initiative on Edge deeP learnIng foR pArticle PHYsics, EPIGRAPHY.