ABSTRACT
In the previous chapters, we generally considered linear structures to perform both supervised and unsupervised tasks. In spite of their interesting trade-off between mathematical tractability and performance, linear structures may be unsatisfactory in some problems, as in the previously discussed case of nonlinear mixtures separation. In fact, a gradual emergence of more complex signal processing scenarios leads to an increasing interest in the use of nonlinear filtering structures, which are able to provide more flexible input-output mappings.
