direkt zum Inhalt springen

direkt zum Hauptnavigationsmenü

Sie sind hier

TU Berlin

Inhalt des Dokuments

Recommendations Through Multirepresentation

Piotr Faliszewski (AGH University of Science and Technology, Krakow)

We introduce a general model for the collective selection of a set of items, where the impact of the value of a selected item on the utility of an agent depends of the rank of this item in the list of selected items ranked according to the agent's preferences. This model encompasses situations where agents benefit from preferred items more than from dispreferred items; it also encompasses proportional representation scoring rules that range between K-winner range voting (where all K members of the selected committee count equally in the utility of an agent) and K-winner Chamberlin-Courant's rule (where only the most preferred one of the K selected candidates counts). We make use of ordered weighted average operators (OWA), and give a few concrete models corresponding to some specific such operators. We show that, in general, these rules are NP-hard to compute and hard to approximate, but-depending on the position of the rule on the spectrum of rules and the type of voters' utilities-there may be good approximation algorithms.

06.02.2014 16:15
Piotr Faliszewski
TEL 512

Back to the research colloquium site.

Zusatzinformationen / Extras