Plenary lecture 7: variable structure-based learning algorithms for neural networks

This plenary speech covers Control schemes for nonlinear dynamical systems using Variable Structure Control (VSC)-based adaptation algorithms for Neural Networks (NN). The VSC approach has been used in diverse control applications, and is new in the NN area. Some of the features of these algorithms are: Finite time convergence to zero of the learning error, guaranteed by stability analysis, robustness with respect to input and external perturbations and easy for computer implantation. The presentation includes identification and control applications in order to illustrate the feasibility of the approach. The plenary speech will contemplate the following topics: - VSC-based algorithms for a single neuron and for multilayer NN. - Dynamical filter-weights Neuron. - Dynamic NN VSC- based Adaptive Control of a Class of unknown Nonlinear Systems. - On-line Identification of a direct and inverse transfer operator for dynamical systems. - Inverse model-based Control using NN with VSC-based adaptation algorithms. - Model Reference Adaptive Control using a VSC-based Neuron-like virtual model. - Further works in the area of VSC-based learning algorithms applications.