- Research Statement on Hierarchical Learning in Computational Intelligence
since 2002-12-29
NOT yet completed
Jie Bao
AI-lab, Computer Science,
Iowa State University
baojie@cs.iastate.edu
1999-11-09 1st version
2001-11-02 2nd version
2002-03-07
3rd version
2002-12-29 4th version
Ecology View of Computational Intelligenceby Jie Bao 2000-12-19 Computational intelligence includes neural network, multi-agent system (MAS), evolution computation and artificial immune system. The basic idea of computational intelligence is ¡°social computation¡±, that¡¯s, complex intelligence can be obtained self-organizingly by simple intelligence individuals under some simple social rules (including competition, cooperation and so on). Such a ¡°social computation¡± system can be regarded as an artificial ecology system, which has similar property and development to nature ecology system. So basic laws of computational intelligence can be regarded as ¡°general ecology¡±. Some general laws in computational intelligence, such as order-increasing; information interchange; hierarchy structure and development; progressive centralization; progressive mechanization, are in fact general properties of a kind of self-organize systems. Therefore, the development of computational intelligence, especially the MAS, is closely related to the development of life sciences and social sciences. Neural Network (top-down) and MAS (bottom-up) are integrated methods to carry out the research in practice. |
Sociobiological Principles in Computational Intelligence(Connectism)by Jie Bao, 2000-05-11 Some basic principles or phenomena in sociobiology such as Kin Selection or ¡°selfish gene¡± hypothesis which can lead to system ordering (e.g. ¡°Evolutionarily Stable Strategy¡±), and methodologies of sociobiology such as game theory, can be found in the computational intelligence system. As a branch of artificial intelligence, computational intelligence includes neural network, genetic algorithms, evolution computing, artificial immune system, and multi-agent system. Its philosophy is ¡°social computing¡±, i.e., complex intelligence can ¡°emergence¡± from competition and cooperation between computing individuals which have only simple intelligence, along with simple social regulations. From the viewpoint of dynamics, and supported strongly by recent research, computational intelligence system and bio-society system are similar. For example, sub-population evolution and symbiosis group selection recent proposed in evolution computing is the embodiment of kin selection; we can also let genetic individuals of different ranks have different search strategies, instead of using classical algorithm that all individuals have the same strategy, to build a food chain in artificial evolution computing system and got desired solution eventually from the evolutionarily stable state; Darwinism methods and game theory¨Cbased negotiation are widely used in multi-agent ecology system. Those all are not accidentally but determined by the basic common property of computing intelligence system and bio-society system of being order-increasing system, or ¡°information system¡±. Based on this philosophy, we proposed Behavior Evolutionary Genetic Algorithm (BEGA) and Hierarchical Neural Network (HNN). BEGA regards ¡°behavior¡± as one kind of evolutionary object which is evolving along with the phenotype feature of the inherit individual. Many kinds of behavior tactics are competing in the evolutionary system and determine the fitness of the individual along with feature. The biological background of BEGA also includes that lot of DNA segments are not corresponding to any protein, and those segments maybe control the behavior of gene and individual. HaNN is based on a philosophy like Kin Selection, and think the competition and evolution of patterns should have the feature of ¡°kin first¡±; in the practice, It uses simple neural networks as basic units to construct complex neural network and allot learning task at global and local levels, and introduces multi-local competition to resolve the problem of weak-robustness and slow-convergence resulted form sheerly global competition. |
Hierarchical learning in Neural Network
Hierarchical learning in Multi Agent System
Hierarchical learning in Genetic Algorithm
Hierarchical learning inNeural Network
Hierarchical learning in Multi-agent System
Hierarchical learning in Genetic Algorithm
[Return to Jie Bao's Homepage]