Evolutionary Computation as a Multi-Agent Search: a $-Calculus Perspective for its Completeness and Optimality

Evolutionary computation in its essense represents a multi-agent competitive probabilistic search. It is useful for solutions of polynomial and hard optimization problems. The solutions found by evolutionary algorithms are not guaranteed to be optimal and evolutionary search is computationally very expensive. Using a generic $-calculus approach to AI, based on process algebras and anytime algorithms, we show that evolutionary search can be considered a special case of $-calculus k -search, and we present some results about completeness, optimality and search costs for evolutionary computation. The main result of the paper is to demonstrate how using $-calculus to make evolutionary computation totally optimal, i.e., how to allow to find the best quality solution with minimal search cost.