Beyond Kemeny Medians: Consensus Ranking Distributions Definition, Properties and Statistical Learning

3 avr. 2026, 16:40
50m
Centre de Conférences Marilyn et James Simons (Le Bois-Marie)

Centre de Conférences Marilyn et James Simons

Le Bois-Marie

35, route de Chartres CS 40001 91893 Bures-sur-Yvette Cedex

Orateur

Ekhine Irurozki (Télécom paris)

Description

Summarising a distribution over rankings by a single Kemeny median fails whenever the distribution is multimodal or heterogeneous. Drawing on the histogram analogy, we introduce Consensus Ranking Distributions (CRD): sparse mixtures of local Kemeny medians indexed by a partition of the space of rankings, interpolating between a single consensus ranking and the raw empirical distribution. We propose the COAST algorithm, a top-down decision tree that learns the partition from data using pairwise comparison splits, and establish a PAC-style generalisation bound. Experiments on synthetic mixtures and real preference data illustrate the method's ability to recover modes and produce interpretable summaries.

Documents de présentation

Aucun document.