BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//CERN//INDICO//EN
BEGIN:VEVENT
SUMMARY:The Dynamics of Information in Neural Networks: From Interacting P
 articles to Continuum Flows
DTSTART:20270117T180000Z
DTEND:20270122T130000Z
DTSTAMP:20261109T120200Z
UID:indico-event-16529@indico.global
CONTACT:contact@swissmaprs.ch
DESCRIPTION:Speakers: Andrea Agazzi\, Borjan Geshkovski\, Philippe Rigolle
 t\n\nDeep neural networks\, and in particular transformer architectures\, 
 have become a cornerstone of modern machine learning\, driving unprecedent
 ed advances in Artificial Intelligence (AI). Yet the mechanisms by which t
 hese models learn and process information through their depth remain poorl
 y understood from a mathematical standpoint.\nThis workshop aims to provid
 e an interface for researchers in machine learning theory and mathematical
  physics/mathematics to investigate the mathematical foundations of inform
 ation processing in deep neural networks\, with a particular focus on tran
 sformer architectures. The natural point of contact between these fields l
 ies in interpreting the evolution of parameters during training and of dat
 a representations during inference as dynamical systems\, where neurons or
  features evolve as particles coupled through interaction terms resulting 
 from the choice of network architecture or training algorithm. In appropri
 ate scaling limits\, these interacting particle dynamics give rise to cont
 inuous flow descriptions capturing the emergent\, macroscopic behavior of 
 deep networks. This connection provides a natural bridge between learning 
 theory and established areas of mathematical physics\, statistical mechani
 cs\, and analysis\, which the workshop aims to strengthen by fostering col
 laborations toward a unified mathematical theory of learning and inference
  dynamics in modern neural network architectures.\nFor more information: h
 ttps://math.mit.edu/~rigollet/Diablerets/\n\nhttps://indico.global/event/1
 6529/
IMAGE;VALUE=URI:https://indico.global/event/16529/logo-2098723454.png
LOCATION:SRS
URL:https://indico.global/event/16529/
END:VEVENT
END:VCALENDAR