Spatial–temporal knowledge base for modeling and analysis of evolving systems

This paper proposes a novel spatial–temporal type of knowledge base that is able to capture the continuously changing spatial and temporal characteristics of a real evolving system, by measuring, processing and aggregating the online data streams coming from the system. By assumption, the real system under investigation is a kind of evolving system that dynamically changes (evolves) its parameters, structure and behavior with time. The proposed spatial–temporal knowledge base normally consists of finite number of memory cells that gradually develop two kinds of memory, namely spatial and temporal memory. The spatial memory aggregates the information in the vicinity of the location of the memory cell, while the temporal memory has a memory buffer with a limited length that keeps the weighted outputs from the real system for further averaging, according to their importance in time. Thus, the continuous (and practically endless) information from the data streams is aggregated and reduced in an efficient way in the knowledge base, so that to keep a relatively small portion of up-to-date information at each sampling period. All computational details of the proposed knowledge base are explained in the paper. As one of the many possible applications, the problem of detection of changes in evolving image systems is illustrated in the paper by use of the proposed double buffer method. Some discussions in the final section of the paper reveal possible future problems to be solved as well as possible application of this computational platform.