Batch SOM algorithms for interval-valued data with automatic weighting of the variables

The Kohonen self-organizing map (SOM) is an unsupervised neural network with a competitive learning strategy that uses a neighborhood lateral interaction function to discover the hidden topological structure of the input data and has both visualization and clustering properties. In this presentation, we propose batch SOM algorithms with automatic weighting of the variables to training the Kohonen network, aiming to cluster interval-valued data while preserving their topology. Applications on synthetic and real interval-valued data sets are proposed aiming to show the usefulness of the proposed batch SOM algorithms. Contact « Séminaire de Statistique appliquée » : Mr Pierre-Louis Gonzalez Cnam – Departement IMATH E.mail : [email protected] Page web consacrée au séminaire: http://maths.cnam.fr/spip.php?article247