Optimal strategies for quantitative data retrieval in distributed database systems

Proposes a new methodology for the optimization of the query processing in relational and table database networks, and its theoretical foundations. As optimization criteria, the author applies the anticipated cost and time of search, as well as the amount of information retrieved. The latter criterion is based on the notion of relative information contents which the author introduces here for relational and table databases. The author formulates the multicriteria optimization problem and derives the optimal sequence of retrieving information from different databases. Optimal search strategies so derived may be regarded as Markov decision trees. The query processing system is supplemented by a learning scheme allowing one to update ex post the knowledge base characterizing the information contents of each database in the network. Finally, a practical implementation of the above system is discussed