BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:Techniques for adaptive machine learning and their applications to
  recommendation systems
DTSTART:20160624T090000
DTEND:20160624T110000
DTSTAMP:20260501T053544Z
UID:be40d91e52dde28ed44cbe14bd47ab5c5a22fec18e488341ddf9560c
CATEGORIES:Conferences - Seminars
DESCRIPTION:Fei Mi\nEDIC Candidacy Exam\nExam President: Prof. Karl Aberer
 \nThesis Director: Prof. Boi Faltings\nCo-examiner: Prof. Volkan Cevher\nB
 ackground papers:Collaborative filtering with temporal dynamics\, by Y. Ko
 renBayesian variable order Markov models\, by C. DimitrakakisA contextual-
 bandit approach to personalized news article recommendation\, by Lihong Li
  et al.Abstract\nMachine learning is often used to acquire knowledge in do
 mains that undergo frequent changes\, such as networks\, social media\, or
  markets. These frequent changes pose a chal- lenge to most machine learni
 ng methods as they have difficulty adapting. In this proposal\, we discuss
  three existing works of modeling temporal factors which shed lights on ou
 r proposal for building adaptive machine learning models. The first work e
 xplicitly models temporal factors for latent factor models and neighborhoo
 d models. The later two works formalize the problem as a sequence predicti
 on problem and propose two online algorithms based on context tree structu
 re and multi-arm bandit formulation respectively. To start with our resear
 ch\, we compare different recommendation methods\, including a new context
  tree (CT) method. The results show that the CT recommender performs bette
 r than other baseline methods due to its adaptation to changes in the doma
 in\, which highlights the importance of considering this aspect when using
  machine learning.
LOCATION:INR 212 http://plan.epfl.ch/?lang=fr&room=INR+212
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR