Advances in Learning Theory: Methods, Models and Applications

New methods, models, and applications in learning theory were the central themes of a NATO Advanced Study Institute held in July 2002. Contributors in neural networks, machine learning, mathematics, statistics, signal processing, and systems and control shed light on areas such as regularization parameters in learning theory, Cucker Smale learning theory in Besov spaces, high-dimensional approximation by neural networks, and functional learning through kernels. Other subjects discussed include leave-one-out error and stability of learning algorithms with applications, regularized least-squares classification, support vector machines, kernels methods for text processing, multiclass learning with output codes, Bayesian regression and classification, and nonparametric prediction.