Martedì 30 gennaio, alle ore 12:00, nell'aula G (ex Psicologia) del Dipartimento di Matematica e Fisica della Università degli Studi della Campania “Luigi Vanvitelli”, il  Professore Francisco de A. T. de Carvalho (Universidade Federal de Pernambuco - UFPE, Recife, Brasil) terrà un seminario dal titolo

«Gaussian Kernel C-Means Clustering Algorithms with Automated Computation of Bandwidth Parameters»

ABSTRACT: Conventional Gaussian kernel c-means clustering algorithms are widely used in applications. However, Gaussian kernel functions have an important parameter, the bandwidthparameter, that needs to be tuned. Usually this parameter is tuned once and for all,  and it is the same for all the variables. In this way, implicitly the variables are equally rescaled and therefore, they have the same importance on the clustering task. This paper presents Gaussian kernel C-Means clustering algorithms with automated computation of bandwidth parameters. In these kernel-based clustering algorithms, the bandwidth parameters change at each iteration of the algorithm, they are different from one variable to another, and they can be different from one cluster to another. Because each variable is rescaled differently according to its own hyper-parameter, these algorithms are able to select the important variables in the clustering process.

Examples with data sets of the UCI machine learning repository corroborate the usefulness of the proposed algorithms.

Il Seminario è proposto dalla prof.ssa Rosanna Verde