This paper demonstrates that holon networks can be used effectively for the identification of large-scale nonlinear dynamical systems. The emphasis of the paper is on modeling of complicated systems, such as complex adaptive systems which bring a new paradigm of science. The concept of applying a quantitative model, for example, to environmental or ecological systems is not new. Recurrent network models are useful since their functions have great flexibility. At the same time, however, such networks are computationally expensive to learn the target functions. In this paper we propose a new recurrent network model of complex systems, called holon network. The evolution of holon networks adapting to environmental systems is achieved by an autonomous decentralized heuristic method based on the concept of holons, which yields a valid relation between whole systems and the constituents. The evolution method with holon's autonomy can reduce the computation cost by adapting the number of holons to environment. Some examples show that the networks organized by this method are able to adapt well their functions to the environmental changes.
ASJC Scopus subject areas