Self-programming network (SPN). A computational model for adaptive evolutionary computers
T. Furuya, Yuki Sato, Hidetaka Ito, T. Higuchi, Yasuhíro Suzuki · 2005
We intend to develop human-like computers which can work in ambiguous, changing, or unpredictable environments. Such computers can acquire knowledge adaptively and evolutionally by interacting with the external environment. A computational model for the computers is proposed. The proposed model is a network of adaptive and evolutional modules. The module consists of neural networks and is based on finite automata. It has a selection and evolution mechanism by which not only the network structure but also the structure of each module are changed adaptively. The program based on this model consists of network structures and rules in each module, i.e. the program is distributed on a network and is self-organized by adaption. Complex behavior or intelligence emerges through the interaction between modules. A distributed memory based architecture for the adaptive evolutionary computer is proposed.