MALOTEC is organizing a seminar with Samir Loudni on Thursday, April 12 at 11:00 am in Room B013.
Samir Loudni is a lecturer at the University of Caen in Normandy, and a member of the GREYC laboratory. He is applying for the position of University Professor at the University of Lorraine.
Title: “Declarative approaches to pattern-oriented data mining”.
Constraint programming (PPC) and integer linear programming (PLNE) provide a generic and flexible framework for solving constraints optimization problems. The use of PPC / PLNE to express data mining problems has many advantages. The first is to offer the user a simple and declarative way to model their problems. The second is to propose a generic resolution approach that allows the user to no longer have to worry about writing a specific algorithm for each search task. In this talk, I will present some contributions regarding the contributions of PPC for pattern-oriented data mining and more specifically on the extraction of sequential patterns and Pareto patterns. Then I will show how PLNE can be exploited to model optimization tasks in data mining, such as the problem of conceptual clustering, or tiling. Finally, I will conclude with our current work on the consideration of the principle of equity to generate balanced clustering using OWA (Ordered Weighted Averages) type collective aggregation functions.