Stochastic versus deterministic concepts in adaptive systems design

Adaptive systems for signal processing can be characterized from both stochastic and deterministic viewpoints. A review of the underlying optimality principles, emphasizing the local aspects of adaptive algorithms is given. Time variance is introduced as an explicit concept of algorithm design; and a solution in terms of a coefficient vector predictor is presented. Finally, the question of robustness of adaptive systems is explored with respect to numerical stability