Neural networks in multidimensional domains. Fundamentals and new trends in modelling and control (Q1271337)

This monography of 165 pages is concerned with multi-layer perceptrons (MLP) in complex, vectorial and hypercomplex domains. The complex multi-layer perceptron (CMLP) has the structure of classical multi-layer perceptrons but over the field of complex numbers. Both analytic and non-analytic activation functions are considered. Vectorial neural networks are multi-input, multi-output extensions of classical single-input, single-output neural networks. They have matrix weights. Then follows MLP in quarternion algebra. The apparatus of the monography becomes more and more boring, too. We suggest that in the next work, the authors should use the exterior calculus. Short time prediction of chaotic time series is applied to Chua and Lorenz systems. Monography ends with applications of quaternions in robotics giving such exotic control of a quaternionic proportional-derivative controller with a feed-forward neural term. Various results are documented by numerical simulations. Matlab routines that implement the basic quaternionic functions are presented.