Abstract:
This paper proposes a theoretical justification and practical implementation in the form of an algorithm for modifying variable-valued unions of functions when adding new production rules to an already formed (within the original subject area) union classifier. The proposed algorithm is based on the application of the method of constructive transformation of variable-valued relationships of classifiers built on the basis of a system of production rules encoded using variable-valued predicates. The properties of changing the structure of production clauses and knowledge development clauses in the process of adding new production rules are studied. Conditions are found under which these clauses are guaranteed to vanish in the connection, or are in a constant form. Taking into account the conditions in the proposed algorithm makes it possible to reduce the number of necessary operations and reduce the computational costs for the required transformations
Keywords:combination of operations, variable-valued predicate, variablevalued logical function, training sample, production rule, mixed classifier, logical neural network.