Apprentissage dans les espaces de grande dimension : Application à la caractérisation de tumeurs noires de la peau à partir d'images
Arthur Tenenhaus · HAL (Le Centre pour la Communication Scientifique Directe) · 2006
The goal of the first part of the thesis is to define concepts allowing developing efficient classification tool for high dimensional data. In the second part of the thesis, we proposed methods based on supervised dimensionality reduction principle. These methods are based on PLS regression which is especially efficient when dealing with high dimensional data. In this context, we proposed a non-linear and binary classification tool, the Kernel Logistic PLS regression. It is based both on latent variables construction and learning with Kernel. We extend the Kernel Logistic PLS regression to the multiclass case, giving rise to the Kernel Logistic Multinomial PLS regression. The two last chapters deal with applications such as melanoma detection from medical imaging or classfication of microarray data.