Evento com Patrice Bertrand ocorre no dia 10 de junho de 2026
O Centro de Informática (CIn) da UFPE receberá Patrice Bertrand, professor da Université Paris-Dauphine – PSL, para apresentação do Seminário de Pesquisa intitulado “On Multilevel Clustering Structures defined as interval convexities, with an application to Machine Learning”. A fala abordará a estrutura de agrupamento hierárquico, um modelo amplamente utilizado em Aprendizado de Máquina, que envolve o agrupamento de objetos em clusters com base em sua similaridade, sem exigir um número predeterminado de clusters. O evento acontecerá no dia 10 de junho de 2026, às 11h, no Anfiteatro do CIn, sem necessidade de inscrição prévia para participar.
Confira abaixo mais informações sobre o seminário e o palestrante:
Título
On Multilevel Clustering Structures defined as interval convexities, with an application to Machine Learning
Resumo
In this talk, we will focus on the hierarchical clustering structure, a widely used model in Machine Learning, that involves the grouping of objects into clusters based on their similarity, without requiring a predetermined number of clusters. We first recall that Hierarchies of sets, and most multilevel clustering structures, have been characterized as convexities induced by interval functions satisfying specific properties, thus giving rise to a unifying framework for characterizing multilevel clustering structures. Here, we show that this unifying framework can be relevant to data mining practice. First, we provide a flexible characterization of hierarchies and weak hierarchies as interval convexities, and using this approach, we characterise the particular structures that are Apresjan hierarchies and Bandelt and Dress weak hierarchies. Furthermore, we introduce a sequence of path-based dissimilarities, which can be computed recursively, and which is decreasing from an arbitrary dissimilarity to its subdominant ultrametric. These path-based dissimilarities define two sequences of nested families of interval convexities. One of the two sequences grows from the Apresjan hierarchy to the single-link hierarchy. An application to the simplification and validation of the single-linkage hierarchy will be illustrated.
Mini-bio
Patrice Bertrand received the Ph.D. degree in Applied Mathematics in 1986 from Université Paris-IX Dauphine, France. He is currently an Associate Professor at Université Paris-Dauphine (France) and a member of CEREMADE, a research centre in Decision Mathematics. He was an external research collaborator at INRIA-Rocquencourt (France) from 1992 to 2013. His research interests focus on ordered sets, theory of cluster analysis, more particularly, clustering structures that extend the classical hierarchical model and allow overlapping clusters, together with cluster validation based on stability measures. He has authored a number of research papers in international journals and conferences. He was a member of several scientific committees of national and international conferences that deal with Data-Mining. He was the scientific secretary of the Société Francophone de Classification (SFC) from 1992 to 1993, the IFCS Council representative of SFC from 2007 to 2010, and the scientific secretary of the International Federation of Classification Societies (IFCS) from 2011 to 2013. He has also served as reviewer of many international journals and conferences.
Comentários desativados