Position:  Researcher at EADS
                in

Applied Mathematics                                 

 Office:  12 rue

Pasteur

             92152 Suresnes, France

   
   
 EMail:  nabil(dot)rachdi@eads(dot)net

                

    

   



                                       

In my PhD thesis, defended at the IMT in December 2011, I have been working on gathering Statistical Learning theory with Complex Systems, seen as black-box functions. As the considered complex input/output models are only known through simulations, I proposed an original statistical framework different from the classical ones which takes into account the simulation aspect. Roughly speaking, instead of having at disposal only observation data, we have in addition simulation data provided from the input/output models that have to be learned. I investigate learning algorithms for parameter estimation in black-box models and we treat this problem as a M-estimation problem. We also study the duality between the parameter estimation procedure and the wanted feature prediction.