Indico "indico.scc.kit.edu" will be now avilable on " indico.kit.edu".

 
Focus on:  Hide Contributions
Back to Conference View
Login

1st collaboration workshop on Reinforcement Learning for Autonomous Accelerators (RL4AA'23)

→ Europe/Berlin
Building 345 (Institute of Beam Physics and Technology (KIT))

Building 345

Institute of Beam Physics and Technology (KIT)

Weingartener Str. 76344 Eggenstein-Leopoldshafen
Andrea Santamaria Garcia (KIT)
Description

Reinforcement learning (RL) is a powerful learning paradigm of machine learning (ML).  It is particularly suited to tackle control problems in large environments, can learn from experience without the need of a model of the dynamics, and can deal with delayed consequences.

Capturing your control problem as a meaningful Markov Decision Process (MDP) is not trivial. Additional challenges arise in the training in terms of stability and evaluation. Other practical aspects include reproducibility, efficiency, implementation, deployment in hardware, or choosing the most suitable algorithm for your problem.

RL applications in particle accelerators are very promising, but have been deployed in real machines only a handful of times. This workshop aims at lowering the barrier in applying RL and making it a more widely used tool.

We will take time for valuable discussions on the topic of RL applied to particle accelerators and to improve our common knowledge. We will also exchange ideas about promising future avenues and help consolidate a community to advance in this area of research in accelerator controls.

Contact
Participants
31 View full list
  • Mon, February 20
  • Tue, February 21
    • 8:30 AM
      KIT-Shuttle to CN Shuttle stop (https://maps.app.goo.gl/YEYs2qa4TdyRmKAR9?g_st=ic)

      Shuttle stop

      https://maps.app.goo.gl/YEYs2qa4TdyRmKAR9?g_st=ic
    • Introduction to RL workshop KARA hall seminar room (https://goo.gl/maps/bWuXfGLFZaRVvhuV8)

      KARA hall seminar room

      https://goo.gl/maps/bWuXfGLFZaRVvhuV8

      Introductory talks and hands-on tutorials

      • 23
        Introduction to RL
        Speaker: Dr Andrea Santamaria Garcia (KIT)
        reinforcement-learning.pdf
      • 24
        Advanced concepts in RL
        Speaker: Dr Simon Hirländer (University of Salzburg)
        Hirlaender_advanced_concepts.pdf
      • 10:40 AM
        Coffee break
      • 25
        Python tutorial: application to an accelerator problem

        Please find the instructions for the tutorial here: https://github.com/RL4AA/RL4AA23

        Speakers: Dr Andrea Santamaria Garcia (KIT), Chenran Xu (IBPT), Jan Kaiser (DESY), Dr Simon Hirländer (University of Salzburg)
    • 12:45 PM
      Lunch break Building 145 (Canteen "Casino")

      Building 145

      Canteen "Casino"
    • 26
      IBPT seminar talk by Dr. Simon Hirländer "Reinforcement Learning in particle accelerator control: are we there yet?" KARA hall seminar room (https://goo.gl/maps/bWuXfGLFZaRVvhuV8)

      KARA hall seminar room

      https://goo.gl/maps/bWuXfGLFZaRVvhuV8
      Speaker: Dr Simon Hirländer (University of Salzburg)
      Hirländer_IBPT_seminar.pdf
    • 3:00 PM
      Coffee break
    • Discussion on advanced RL topics
      • 27
        Discussion Group 1: Modeling, limitations, and methods
        Speaker: Simon Hirländer
      • 28
        Discussion Group 2: Challenges I
        Speaker: Chenran Xu (IBPT)
      • 29
        Discussion Group 3: Challenges II
        Speaker: Michael Schenk
      • 30
        Discussion Group 4: Community
        Speaker: Jan Kaiser
      • 31
        Farewell
        Speaker: Andrea Santamaria Garcia (KIT)
Powered by Indico v3.3.6