The chapter is devoted to the use of predicate calculus for artificial intelligence (AI) problem solving. Here, an investigated object is represented as a set of its elements and is characterized by a fixed number of predicates. Its description is a set of all constant literals (with the chosen predicates), which are valid on the object. The NP-complete problem, “whether an object satisfies a goal formula,” is under consideration. The upper bound of number of its solution steps is exponential. The notion of common up to the names of arguments subformula of two predicate formulas and one of their isomorphisms allows to construct a level description of the set of goal formulas and essentially to decrease the upper bounds of the problem solving. The level description permits to define a self-training predicate network, which may change its configuration during the process of training. The extraction of common up to the names of arguments subformulas permits to construct a multiagent description of an object when every agent does not know the true number of the object elements and uses her own notifications for the names of elements. A model example illustrating all algorithms is presented.
Part of the book: Intelligent System