Vers la découverte de modèles exceptionnels locaux : des règles descriptives liant les molécules à leurs odeurs
In EGC 2015, vol. RNTI-E-28, pp.305-316
Following a complex phenomenon starting with an odorant molecule to the perception in the brain, olfaction is the most difficult sense to understand by neuroscientists. The main challenge is to establish rules on the physicochemical properties of molecules (weight, number of atoms, etc.) to characterize a specific subset of olfactory qualities (fruity, woody, etc.). Subgroup discovery make it possible to find such descriptive rules. However, existing methods provide characterization of either a single label or all the label (exceptional model mining). We then propose an approach for discovering subgroups that characterize only some labels with a new enumeration technique, stemming from redescription mining. We then evaluated this method on an olfactory database provided by the neuroscientists by comparing it with the state-of-the-art algorithm.