Supervisory authorities

CNRS Dauphine PSL *



20 March 2018: One event


  • Séminaires du Pôle 3 : "Sciences des données"

    Tuesday 20 March 11:00-12:00 - Benoît Gaüzère - INSA de Rouen/LITIS

    Séminaires du Pôle 3 : "Sciences des données"

    Résumé : Graphs allow to encode structural information included within data used in chemical or pattern recognition problems. However, conversely to vectors defined in an euclidean space, the definition of a graph (dis)similarity measure is not straightforward, but required to compute prediction models. One of the most well known dissimilarity measure is the graph edit distance. Despite its good interpretability, the computation of a graph edit distance between two graphs is an NP-Hard
    problem. Therefore, its application remains limited to small graphs. During this presentation, I will introduce a formal definition of this metric between graphs as a quadratic assignment problem and some methods used in pattern recognition to approximate an optimal solution. Considering approximations allows us to apply this framework to chemoinformatics problems.

    En savoir plus : Séminaires du Pôle 3 : "Sciences des données"