Titre : Multiclass classification for S.D.E. paths
In this talk, I will present the multiclass classification problem for functional data. In particular, I will consider the case where the features come from a mixture of time-homogeneous diffusion processes. Specifically, the classes are discriminated by the drift function while the diffusion coefficient is common to all classes and unknown. Two cases will be discussed, easier the drift functions have a parametric shape and only a parameter is unknown, the drift functions are totally unknown. We will see how do derive a rate of convergence for a plug-in type classifier in the second case.