BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:Nonequilibrium Physics in Living Systems and Machine Learning
DTSTART:20221017T161500
DTEND:20221017T171500
DTSTAMP:20260511T073003Z
UID:c5e02aa7b211ea4d94e1805166b131563c25c008a60256218958882c
CATEGORIES:Conferences - Seminars
DESCRIPTION:Dr Yuhai TU\, IBM research Center\, Yorktown Heights\, NY USA
 \nComplex networks from biochemical networks to artificial neural networks
  perform intricate biological functions and machine learning tasks. Most o
 f these complex networks operate far out of equilibrium where equilibrium 
 statistical mechanics fails. In this colloquium\, I will describe our rece
 nt work applying concepts and methods from nonequilibrium physics to biolo
 gy and machine learning. \nPart I:Energy-accuracy tradeoff in biological 
 systems. For a wide range of systems\, we show that the energy cost sets a
 n upper bound for the performance of the intended biological functions. \
 nPart II: Learning dynamics in deep-nets. Our recent study revealed a robu
 st inverse relation between the weight variance in stochastic gradient des
 cend (SGD) and the loss landscape flatness opposite to the fluctuation-res
 ponse relation in equilibrium systems.
LOCATION:CM 1 https://plan.epfl.ch/?room==CM%201%201
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR