Automated recurrent neural network design to model the dynamics of complexsystems

A general purpose implementation of the tabu search metaheuristic, called U niversal Tabu Search, is used to optimally design a locally recurrent neura l network architecture. The design of a neural network is a tedious and tim e consuming trial and error operation that leads to structures whose optima lity is nor guaranteed In this paper, the problem of choosing the number of hidden neurons and the number of taps and delays ill the FIR and IIR netwo rk synapses is formalised as an optimisation problem, whose cost function t o be minimised is the network error calculated on a validation data set. Th e performance of the proposed approach has been tested on the problem of mo delling the dynamics of a non-isothermal, continuously stirred tank reactor , in two different operating conditions: when a first order exothermic reac tion is occurring; and when two consecutive first order reactions lend to a chaotic behavior. Comparisons with alternative neural approaches are repor ted showing the usefulness of the proposed method.