A simplification of the theory of neural groups selection for adaptive control

Abstract

Mathematical models have been extensively used to model living organisms behaviour. Nonetheless, those models do not take into account the role that individual history plays into the establishment of neural structures responsible for its interaction with the environment. TNGS has been postulated as a global, comprehensive solution that models individual behaviour from both biological and evolutionary aspects. Besides, it provides a non symbolic approach to learning processes that does not require extensive prior knowledge from system designer.

This paper presents a simplification of TNGS oriented towards its use in adaptive control processes for chemical reactors. An oculomotor system has been implemented based on Darwin III automaton. Simplifications are made on the state equation that describes the dynamic behavior of every processing element. They are driven by a chaotic study on the equation for the weight modification. Based on very simple assumptions (“seeing is better that not seeing”), the system learns to trace a randomly moving object within its vision field. Simulation present data obtained under different assumptions. The evolution of the distance between the center of the input image and the center of the Visual Retina is displayed.

This work has been partially supported by BRITE/EURAM project n∘ BE-7686 PSYCHO (Powerful SYstems for identification and Control of Highly nOn-linear processes using neural networks).