Research And Application On Flexible Hyper Topology Space Model Of Generalized Intelligent System

This paper came from Ph. D foundation program of Ministry of Education, Research on Basic Theories (Universal Symbolism) of Artificial Intelligence (98069923). and national 863 program, Attemper Systems of Produce Plan and Real Time Optimization (2002AA412020).The initial object for the birth of Artificial Intelligence is to use computers to simulate the hominine Abstract Ideation, which is a narrow sense viewpoint. Nature Intelligences exist in each levels of life phenomena widely, such as the hominine Abstract Ideation and Visual Ideation, the creatural learning and intelligent responses, the biologic evolution and immunity, the works-devided and cooperation of biology colonies and the balance of biogeocenose, etc, which contribute to find and epurate some principles and techniques of Artificial Intelligence importantly. At present, the Artificial Intelligence has moved to the viewpoint of generalized intelligence with simulating the Nature Intelligences from the viewpoint of narrow sense.Generalized Artificial Intelligence includes many methods for simulating intelligences, such as the symbolism of simulating Abstract Ideation, the connectism to simulating Visual Ideation, the behaviorism of simulating intelligent behaviors, the multi-agents system of simulating social work-devided and cooperation, the evolution and immunity computing of simulating biologic evolution, and the zoology computing of simulating biogeocenose, etc. Although those have advantages each other and have been applied into different domains successfully, everyone has limitations after all. Research shows that it will get better effect to integrate the above methods properly in solving a material problem on simulating intelligences. So, research on Intelligent Hybrid Systems has become very important research fields in Artificial Intelligence. In the above, the integration between symbolism and connectism is most broad especially. This paper discusses the integration between symbolism and connectism mainly with the direct of Generalized Artificial Intelligence, but its principles and techniques may be extended into the integration of other Artificial Intelligences.Although research on Intelligent Hybrid Systems has gained many fruits, its theories are very imperfect, especially in the partial system models, the limitations of knowledge representation, the oneness of logic neurons and the limitations of knowledge acquisition and problem-solved. This paper is based on the essence ofhuman's intelligent behaviors with an eye to the physiology basis of knowledge representation and storage in one's brains and human's thinking manners, combines the theory of Hyper Topology Space and Universal Logics Principle to exert advantages each other fully, then develops the research work on the Intelligent Hybrid Systems and gets some harvests which are one model, two theories and three technologies with the center of integrations. They are as follows:(1) Flexible Hyper Topology Space ModelThe Flexible Hyper Topology Space Model is a new one of Intelligent Hybrid Systems which is a generalized problem-solved model based on the essential rules of thinking.(2) Hyper Topology Space TheoryThe theory is a more integrated one which is formed elementarily according to the essential rules of thinking. It can describe the frame of knowledge storage and processing in one's brain properly, then establish the basis for advanced research on all sorts of thinking rules and its correlations.(3) Universal Logics NeuronsThe Universal Logics Neurons are better ones for simulating the Abstract Ideation. They have great generality which embodies the flexible rules of Abstract Ideation.(4) Self-organized learning techniqueThat's a self-organized learning technique with nesting, recursion and changeable structure. It's based on the knowledge representation of hyper topology. exerts equivalence relations to seek equivalence classes for mining new knowledge which is contained or connoted, then gets and updates knowledge automatically.(5) New techniques of rules extraction and insertionThose techniques of rules extraction and insertion are simple, visual and effective. They can obtain and refine the rules with special semantics and structures automatically, and be in favor of discovering new knowledge.(6) Unified reasoning and searching techniques based on instructions set Those include a reasoning technique with values flow and a searching techniquewith markers which are presented by simulating parallel and effective thinking. They are based on hyper topology structure graphs and have unified instructions set. In addition, they can be operated simply and visually and the results are easy to be understood, and are flexible to be able to apply broadly.Those above fruits are supplement each other. The model is a frame, the theoriesare basis, and the techniques can be seen as bridges, which are all necessary to construct the building of Intelligent Hybrid Systems altogether so as to infuse new energies for the development of Artificial Intelligence.Lots of instances and the applications in recognition of Chinese characters show that the above fruits are correct and effective.At last, this paper prospects the developing trends of Intelligent Hybrid Systems. and points out the advanced research objects.