Research

Research interests

• Model-based clustering
• Parcimonious Gaussian mixture modelsModèles de mélanges gaussiens parcimonieux
• CEM clustering algorithm (Classification, Estimation, Maximisation)
• Choice of the model and of the number of components
• Block clustering
• Clustering of binary and categorical data
• On-line clustering

Collaborations

• Pôle régional Diva: on-line clustering and clustering of binned data
• Cetim: software for real time flaw diagnosis application on pressure equipments using acoustic emission control
• Renault: clustering of customer satisfaction survey data

Students

• Previous: M. Nadif, F. Marchetti, Y. Bencheikh, S. Elie, C. Ambroise, C. Biernacki, M. Dang, N. Teuscher, M. Bzoui, G. Putz, A. Bessadok, A. Samé, H. Hamdan, C. Gouget
• Current: P. Gaillard, F. Chamrouki, J. Caudeville, J.-B. Masson

Software

• MIXMOD fits mixture models (Gaussian, Bernoulli, multonomial) to a given data set with either a clustering or a discriminant analysis purpose. It is publicly available under the GPL license and is distributed for different platforms (Linux, Unix, Windows). It is developed jointly by INRIA futurs (action SELECT), the laboratory of mathematics of Besançon and the Heudiasyc laboratory of Compiègne.
• Software for spatial clustering
• LOTERE (with Cetim) : software for real time flaw diagnosis application on pressure equipments using acoustic emission control