The Most Simple Systolic Architectures for Parallel Probabilistic Searching Algorithms

Abstract

The paper offers a short survey and a systemization of some author’s results concerning the computational architectures based on systolic arrays with stochastic communication and co-operation among the processors and proposed as tools for implementation of parallel probabilistic searching algorithms. Such architectures offer some possibilities to develop non-traditional control tools and theory corresponding to them. Searching algorithms also emerge as important subalgorithms in a number of more sophisticated algorithms of artificial intelligence, including those for identification and signal processing. Because of the very limited extent of this contribution only the main ideas are explained and the most simple architectures are presented, namely those without any hierarchical structures and those based on two-level hierarchies. On the other hand, these architectures and the results concerning them are presented within the framework of a new general mathematical model which can be easily extended also to more sophisticated architectures, e.g., to many-level hierarchies. For the same reason, i.e., for the lack of space, the proofs of the presented assertions are replaced by references to proofs of corresponding or similar assertions in more detailed author’s papers.