BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//CERN//INDICO//EN
BEGIN:VEVENT
SUMMARY:Reinforcement learning and its applications at CERN
DTSTART:20260723T113000Z
DTEND:20260723T133000Z
DTSTAMP:20260920T164300Z
UID:indico-event-1690268@indico.cern.ch
DESCRIPTION:Speakers: Matteo Bunino (CERN)\n\nAbstract:\nReinforcement Lea
 rning (RL) has emerged as a powerful paradigm in artificial intelligence a
 nd has found exciting applications in various fields\, including particle 
 accelerators at CERN. This introductory lecture aims to provide an overvie
 w of RL and its application in optimizing beam steering in the AWAKE beaml
 ine.\nThe lecture will begin with an introduction to the fundamentals of R
 L\, where we will explore the concept of agents learning to make decisions
  through interaction with an environment. Central to RL is the concept of 
 Markov Decision Processes (MDPs)\, which model sequential decision-making 
 problems. We will discuss the components of an MDP\, including states\, ac
 tions\, rewards\, and transition probabilities.\nNext\, we will delve into
  sample-based methods\, such as Monte Carlo and Temporal Difference (TD) l
 earning\, which are essential for learning in uncertain and dynamic enviro
 nments. These techniques allow RL agents to estimate value functions and i
 mprove their policies through experience.\nModel-based RL will be introduc
 ed as an approach to learn a model of the environment to aid in decision-m
 aking. Dyna\, a well-known model-based RL algorithm\, will be presented as
  an example of how this integration can be achieved.\nFunction approximati
 on techniques applied to both parametric value functions and policies will
  be covered. These methods enable RL agents to handle high-dimensional sta
 te and action spaces\, making them valuable tools in complex applications.
 \nThe lecture will conclude by introducing actor-critic methods\, which co
 mbine the advantages of both policy gradient and value-based methods. We w
 ill briefly discuss how these algorithms facilitate efficient learning and
  convergence in RL tasks.\nFinally\, we will transition to the exciting RL
  applications at CERN. \nShort bio:\nMatteo Bunino earned a MSc in Data S
 cience and Computer Engineering from both the Polytechnic University of Tu
 rin (Italy) and EURECOM (France). He did his thesis at Huawei's Munich Res
 earch Center (MRC)\, where he developed a prototype for analyzing dynamica
 lly evasive malware with the aid of reinforcement learning.\nAfter univers
 ity\, Matteo joined CERN\, where he worked on interTwin\, a European proje
 ct aimed at developing a unified digital twin engine (DTE) for science\, a
 nd contributed to CERN openlab. During this time he focused on the develop
 ment of "itwinai"\, a framework for advanced MLOps on cloud and HPC.\nCurr
 ently\, Matteo is part of the Kubernetes platform team in the CERN IT depa
 rtment\, where he works on the ODISSEE project\, whose goal is to create t
 he digital twin of the LHCb datacentre.\n\nhttps://indico.cern.ch/event/16
 90268/
LOCATION:31/3-004 - IT Amphitheatre (CERN)
URL:https://indico.cern.ch/event/1690268/
END:VEVENT
END:VCALENDAR