Multiobjectivization by Decomposition of Scalar Cost Functions

摘要: The term `multiobjectivization' refers to the casting of a single-objec-tive optimization problem as multiobjective one, transformation that can be achieved by addition supplementary objectives or decomposition original objective function. In this paper, we analyze how multiobjectivization decompositionchanges fitness landscape given and affects search. We find has only one possible effect: introduce plateaus incomparable solutions. Consequently, hillclimbers using no archive `see' smaller (or at most equal) number local optima on transformed compared problem. When archived are considered effect may partly reversed. Running time analyses conducted four example functions demonstrate (positive negative) influence both itself, use vs. non-use an archive, have performance simple hillclimbers. each case exponential/polynomial divide is revealed.