IRIS-HEP Fellow: Vesal Razavimaleki



Fellowship dates: July - September 2020
Home Institution: University of California, San Diego


Adapting GNN Tracking for FPGAs with hls4ml

Graph neural networks (GNNs) have demonstrated promise for pattern recognition problems like particle tracking. To meet the demands of the planned HL-LHC, there has been increased interest in accelerating large machine learning (ML) models with FPGA coprocessors for integration into the L1 trigger. Deployment of neural networks on FPGAs has been studied with the hls4ml compiler package which uses high-level synthesis to convert ML models to FPGA firmware. This project proposes to expand the hls4ml toolkit to support GNNs for particle tracking, allowing them to be implemented in FPGA coprocessor applications possibly including the L1 trigger.

More information: My project proposal

Mentors:
  • Javier Duarte (University of California, San Diego)



Contact me: