超分布超并行智能处理的同步均质竞争行波方法

SYNCHRONOUS-HOMOGENEOUS PROPAGATION OF CONCURRENT COMPETITIVE WAVES FOR HYPER-DISTRIBUTED HYPER-PARALLEL INTELLIGENT PROCESSING

本文提出的竞争行波的原理、机制和算法，用于超分布超并行智能处理，克服了传统的符号逻辑单行算法和通常的神经网络方法在人工智能问题求解方面面临的许多困难．本文集中论述了竞争行波中最基本的一类同步均质竞争行波，及其超分布超并行问题求解算法，讨论了它们的性质和应用，并与其它著名的经典搜索算法相比较，表明了本文方法多方面的优点及广泛的应用前景．

The artificial intelligence algorithms based on the classical sequentialsymbolic logic has been confronting tremendous difficulties in respect of the real-time performance, the complexities of time and spaces, and hardware implernenta-tion. Although the neural networks' approach has been used in tnatching problemand the optimization,it is still very hard to use heuristic knowledge during search-ing an implicit AND/OR graph by means of neural networks. In order to overcomemany formidable drawbacks present in both approaches of the sequential symboliclogic and the neural networks of the artificial intelligence problem solving,this pa-per proposes the principle of concurrent cotnpetitive wave propagation and the ap-proach to hyper-distributed hyper-parallel intelligent processing. only themost fundarnental case, i. e. the synchronous homogeneous propagation of concur-rent competitive waves is considered in this paper. The basic conception, mechani-sm, concurrent algorithms and their properties, and some application issues are dis-cussed. Through the comparison with other famous traditional algorithms of artifi-cial intelligence, many advantages of the proposed approach, such as the real-timeperformance, the knowledge-based processing in a hyper-distributed hyper-parallelenvironment, the feasibility for hardware implementation and various applications,and so on, are shown. Since searching an implicit AND/OR graph is one of themost important methods in the AI areas, a great attention is paid to the