ABSTRACT
This chapter provides information on a theoretical study of the network which specifies important global properties of the network. It utilizes nonlinear system synthesis and identification methods of Volterra-Wiener series expansion to better predict the system behavior and demonstrate the computational power embedded in multiplicative lateral inhibitory interaction. Linear mathematical methods of system identification are incapable of satisfactorily identifying nonlinear systems. The nonlinear analysis technique of Volterra-Wiener expansions has been successfully applied to, among others, the analysis of wave propagation in random media, scattering by random surfaces, Brownian motion, theory of turbulence, and biological systems. The chapter then presents an analysis of temporal and spatial characteristics of neural networks. Adaptive resonance architectures are neural networks that produce stable recognition codes in response to arbitrarily complex sequences of input patterns and do so by self-organizing, i.e., without requiring an external teacher.
