Statistical Seminar: Every Monday at 2:00 pm.
Time: 2:00 pm – 3:15 pm
Date: 14th March 2024
Place : 3001

Emmanuel PILLIAT (Université Montpellier) – Ranking the rows of a Permuted Isotonic Matrix Optimally and in Polynomial Time

Abstract:

In the context of crowdsourcing, a natural question is how accurately we can determine the ranking of experts (or workers) who are labelling some data. The same kind of question arises in tournaments, where we might want to rank players based on the outcomes of games between pairs of players. We will consider a ranking problem where we have noisy observations from a matrix with isotonic columns whose rows have been permuted by some permutation π*. This encompasses many models, including crowd-labeling and ranking in tournaments by pairwise comparisons. After the introduction of the statistical model and of the risk measures, we will discuss the ideas for an optimal and polynomial-time procedure for recovering π*, settling an open problem in Flammarion et al. (2019). The presented approach is based on iterative pairwise comparisons by suitable data-driven weighted means of the columns.

Organizers:
Anna KORBA (CREST), Karim LOUNICI (CMAP) , Jaouad MOURTADA (CREST)

Sponsors:
CREST-CMAP