Choose timezone
Your profile timezone:
Special EPE Seminar: Elham E Khoda
→
America/Los_Angeles
Zoom-only
Zoom-only
Description
Title:
Journey of Machine Learning at the LHC: Offline to Online
Speaker:
Elham E Khoda
University of Washington
Abstract:
Machine Learning (ML) is creating a revolution in science and technology. Experimental particle physics is increasingly implementing ML developments. I will highlight the ML movements in experimental particle physics by focusing on two research projects.
The process began with the adoption of ML algorithms in place of rule-based ones for offline reconstruction. I will highlight one such application used for offline track reconstruction. Highly collimated charged particle tracks coming from extremely energetic particles frequently produce overlapping responses in the ATLAS detector. I will discuss how these fused charge clusters on the pixel detector can be split efficiently using a Mixture Density Network (MDN)-based algorithm. While ML-based algorithms are powerful for offline event reconstruction, a limitation comes from the data collection step. The particle detectors around the LHC ring use an electronic hardware "trigger" system to select potentially interesting particle collisions for further analysis. Currently, one out of 400 proton-proton collision events passes the hardware trigger. As the collision rate will increase by 5-7 times in the future alternative algorithms, such as ML, can be used for fast and accurate decisions.
In the later part of the talk, I will highlight the potential applications of ML for hardware (ASIC or FPGA) triggers. I will discuss a method to implement the ML algorithms on an FPGA using the hls4ml software package. hls4ml is a user-friendly software based on High-Level Synthesis (HLS) and designed to deploy neural network architectures on FPGAs. Afterward, I will highlight my recent work on recursive neural networks (RNN)-based algorithms for trigger applications.
Biography:
Elham E Khoda completed his Ph.D. in Physics at the University of British Columbia, Vancouver, Canada, on “Searches for new high-mass resonances in top-antitop and di-electron final states using the ATLAS detector” and joined the EPE group as a postdoc working on BSM searches and ML research. He is leading the resonance search in the top-antitop final state pp collisions in the ATLAS detector and actively working in ATLAS tracking in a dense environment. He is a major contributor to EPE's activities toward data-driven discovery with accelerated AI algorithms. He is working on accelerating ML inference with coprocessors like GPUs and FPGAs. He is the current ATLAS Exotics liaison to the ATLAS Machine Learning Forum and co-convener of the Coprocessor group of Fast ML collaboration.
