BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:BSU Seminar: &quot\;A unifying framework for generalised Bayesian 
 online learning in non-stationary environments&quot\; - Gerado Duran-Marti
 n\, Oxford-Man Institute\, University of Oxford 
DTSTART:20260407T130000Z
DTEND:20260407T140000Z
UID:TALK245395@talks.cam.ac.uk
CONTACT:Alison Quenault
DESCRIPTION:We propose a unifying framework for methods that perform prob
 abilistic online learning in non-stationary environments. We call the fram
 ework BONE\, which stands for generalised (B)ayesian (O)nline learning in 
 (N)on-stationary (E)nvironments. BONE provides a common structure to tackl
 e a variety of problems\, including online continual learning\, prequentia
 l forecasting\, and contextual bandits. The framework requires specifying 
 three modelling choices: (i) a model for measurements (e.g.\, a neural net
 work)\, (ii) an auxiliary process to model non-stationarity (e.g.\, the ti
 me since the last changepoint)\, and (iii) a conditional prior over model 
 parameters (e.g.\, a multivariate Gaussian). The framework also requires t
 wo algorithmic choices\, which we use to carry out approximate inference u
 nder this framework: (i) an algorithm to estimate beliefs (posterior distr
 ibution) about the model parameters given the auxiliary variable\, and (ii
 ) an algorithm to estimate beliefs about the auxiliary variable. We show h
 ow the modularity of our framework allows for many existing methods to be 
 reinterpreted as instances of BONE\, and it allows us to propose new metho
 ds. We compare experimentally existing methods with our proposed new metho
 d on several datasets\, providing insights into the situations\n that make
  each method more suitable for a specific task.
LOCATION:Large Seminar Room\, East Forvie Building\, Forvie Site Robinson 
 Way Cambridge CB2 0SR.
END:VEVENT
END:VCALENDAR