A study on efficient searching for image information in large-scale network.

The information possessed by many individual users connected to a large-scale network can be considered as a gigantic database, termed a superdistributed database. To date, effective methods of searching scattered information in such a superdistributed environment have been proposed. The problems are different between the cases where the number of information items to be searched for in the network is 1 or unknown. For both cases, the method proposed in this paper can measure the search effect in terms of evaluation quantity taking into account the communication cost and the time cost. Accordingly, the search considered most effective at that time point is repeated. In this paper, the optimum nature of the proposed search method when either the communication cost or the time cost is taken into consideration is proven. In addition, the characteristics of this method are analyzed for general cases and it is shown that the communication cost and the time cost can be reduced. Next, the effect of the error in the existence probability of the information to be searched at each node in the network is discussed. It is shown that the effect is small in the proposed method. Further, an approximate search procedure with a reduced amount of computation is proposed. It is demonstrated that the effect obtained is similar to those in the search methods presented to date. © 2000 Scripta Technica, Electron Comm Jpn Pt 1, 83(7): 15–25, 2000 [ABSTRACT FROM AUTHOR]