Multiobjective optimization with messy genetic algorithms
David A. Van Veldhuizen, Gary B. Lamont · 2000
This paper extends the traditional notion of building blocks as found in single-objective Evolutionary Algorithms (EAs) to the Multiobjective Optimization Problem (MOP) domain.This effort aims to develop more effective and efficient Multiobjective EAs (MOEAs) by using relevant MOP concepts and a consistent MOEA-Pareto notation to depict building block processing in solving MOPs.Specifically, an innovative extension of the building block-based messy genetic algorithm (called the MOMGA) is presented and applied successfully to our MOP test suite.This approach is shown to be quite effective when compared to other contemporary MOEAs.