Ranking recommendations via online learning
Διατεταγμένες συστάσεις μέσω άμεσης μάθησης
View/
Abstract
These days, that can be characterized by high rates of information transmission and the increasingly usage of world wide web, there are plenty of applications and websites that handle recommendations systems, in order to provide their users with the best possible suggestions. The solution to such systems is given by Online Learning, which demand low computational resources and yield optimum results. In this work, we examine Online Learning and especially the Multi-armed Bandit problem, which is one of its aspects. In later sections, we will present some of the most known online learning algorithms that refer to that problem. Specifically, we are interested in rankings of recommendations and the usage of the meta-algorithm Ranked Bandits Algorithm (RBA) in our experiments, which manipulates instances of other online algorithms. We study the convergence of RBA through “sponsored recommendations”, in which each document is related to an income, while each user is a subset of the documents he is interested in and derives randomly from a set of different populations.
Postgraduate Studies Programme
Πληροφοριακά Συστήματα και ΥπηρεσίεςDepartment
Σχολή Τεχνολογιών Πληροφορικής και Επικοινωνιών. Τμήμα Ψηφιακών ΣυστημάτωνNumber of pages
51Language
EnglishCollections
Except where otherwise noted, this item's license is described as
Αναφορά Δημιουργού-Μη Εμπορική Χρήση-Όχι Παράγωγα Έργα 3.0 Ελλάδα
Αναφορά Δημιουργού-Μη Εμπορική Χρήση-Όχι Παράγωγα Έργα 3.0 Ελλάδα