ADAPTIVE EVOLUTION OF HOLON NETWORKS BY AN AUTONOMOUS DECENTRALIZED METHOD

This paper demonstrates that holon networks can be used effectively fo r the identification of large-scale nonlinear dynamical systems. The e mphasis of the paper is on modeling of complicated systems, such as co mplex adaptive systems which bring a new paradigm of science. The conc ept of applying a quantitative model, for example, to environmental or ecological systems is not new. Recurrent network models are useful si nce their functions have great flexibility. At the same time, however, such networks are computationally expensive to learn the target funct ions, In this paper we propose a new recurrent network model of comple x systems, called holon network. The evolution of holon networks adapt ing to environmental systems is achieved by an autonomous decentralize d heuristic method based on the concept of holons, which yields a vali d relation between whole systems and the constituents. The evolution m ethod with holon's autonomy can reduce the computation cost by adaptin g the number of holons to environment. Some examples show that the net works organized by this method are able to adapt well their functions to the environmental changes. (C) 1998 Elsevier Science Inc. All right s reserved.