Tecniche Di  Machine Learning Con Dispositivi FPGA  per Gli Esperimenti Di Fisica Delle Particelle

Europe/Rome
AULA BP- 2B (Infn)

AULA BP- 2B

Infn

Bologna, Viale Berti Pichat 6/2
Description

The course aims to provide the state of the art on the implementation of Machine Learning (ML) and Deep Learning (DL) techniques in FPGA-type devices in particle physics applications, and to help spread the related know-how. Within INFN, also contributing to increase it thanks to contributions from experts in the field outside the institution.

This course aims to illustrate how ML techniques can be implemented in FPGA devices from a dual point of view. First, the course will focus on the technological aspect: the software and hardware tools available on the market will be presented and development methodologies will be illustrated, discussing practical cases. Secondly, the course will show the state of the art of ML applications on FPGAs within the particle physics community through the analysis of concrete cases and examples of applications (of varying complexity). The course will stimulate and encourage discussion among participants on the current advantages and disadvantages and on the future potential of the different technologies and on the different tools in relation to the individual use cases considered.

Participants
  • Antonio Sidoti
  • Daniele Battista
  • Elena Pedreschi
  • Enrico Calore
  • Francesca Del Corso
  • Francesco Chiapponi
  • Giacomo Levrini
  • Gianluigi Chiarello
  • Giulio Bianchini
  • Ian Postuma
  • Lisa Zangrando
  • LORENZO VALENTE
  • Moh Rafik
  • Nicolò Tosi
  • orlando angelo
  • Paolo Andreetto
Segreteria Locale
    • 14:15 15:00
      Efficient Machine Learning in High-Energy Physics 45m
      Speaker: Jennifer Ngadiuba (CMS)
    • 15:00 15:35
      Introduction to FPGA 35m
      Speaker: Riccardo Travaglini (Istituto Nazionale di Fisica Nucleare)
    • 15:35 16:00
      Deep Learning inference with FPGA 25m
      Speaker: Riccardo Travaglini (Istituto Nazionale di Fisica Nucleare)
    • 16:00 16:30
      Coffee break 30m
    • 16:30 17:05
      Fast inference with HLS4ML: Machine Learning with FPGA at LHC 35m
      Speaker: Thea Aarrestad (ETH Zurich)
    • 17:05 17:55
      Xilinx ML/AI: technologies (edge, on-premise, on cloud) and solutions (VitisAI, Pynq, …) 50m
      Speaker: Mario Ruiz-Noguera (AMD)
    • 09:30 13:00
      HLS4ML tutorial part 1 3h 30m

      Functionalities, quantization and compression

      Speakers: Sioni Summers, Thea Aarrestad (ETH Zurich)
    • 14:30 17:30
      HLS4ML tutorial part 2 3h

      Demo on real HW or deploying on AWS F1

      Speakers: Marco Lorusso (Istituto Nazionale di Fisica Nucleare), Sioni Summers, Thea Aarrestad (ETH Zurich)
    • 09:00 10:00
      Deep Learning inference with the Bond Machine project: introduction and demo 1h
      Speaker: Mirko Mariotti (Istituto Nazionale di Fisica Nucleare)
    • 10:00 11:00
      Open discussion and demo 1h
      Speaker: Riccardo Travaglini (Istituto Nazionale di Fisica Nucleare)
    • 11:00 11:30
      Coffee break 30m
    • 11:30 12:30
      ML/AI with Intel FPGA 1h
      Speaker: Vladimir Loncar
    • 12:30 13:00
      Alternative solutions at the Edge: Lattice, Microchip 30m
      Speaker: Riccardo Travaglini (Istituto Nazionale di Fisica Nucleare)
    • 14:00 14:20
      Introduction to Vitis AI 20m
      Speaker: Riccardo Travaglini (Istituto Nazionale di Fisica Nucleare)
    • 14:20 16:30
      Vitis AI tutorial/hands-on 2h 10m
      Speaker: Marco Lorusso (Istituto Nazionale di Fisica Nucleare)