Conference Paper

Using gradient-based information to deal with scalability in multi-objective evolutionary algorithms

Dept. de Comput., CINVESTAVIPN, Mexico City
DOI: 10.1109/CEC.2009.4982925 Conference: Evolutionary Computation, 2009. CEC '09. IEEE Congress on
Source: IEEE Xplore

ABSTRACT This work introduces a hybrid between an elitist multi-objective evolutionary algorithm and a gradient-based descent method, which is applied only to certain (selected) solutions. Our proposed approach requires a low number of objective function evaluations to converge to a few points in the Pareto front. Then, the rest of the Pareto front is reconstructed using a method based on rough sets theory, which also requires a low number of objective function evaluations. Emphasis is placed on the effectiveness of our proposed hybrid approach when increasing the number of decision variables, and a study of the scalability of our approach is also presented.

0 Bookmarks
 · 
74 Views
  • [Show abstract] [Hide abstract]
    ABSTRACT: This work presents the main aspects to tackle when designing memetic algorithms using gradient-based local searchers.. We address the main drawbacks and advantages of this coupling, when focusing on the efficiency of the local search stage. We conclude with some guidelines and draw further research paths in these topics.
    Genetic and Evolutionary Computation Conference, GECCO 2010, Proceedings, Portland, Oregon, USA, July 7-11, 2010, Companion Material; 01/2010
  • [Show abstract] [Hide abstract]
    ABSTRACT: Most real world multi-objective problems (MOPs) have a complicated solution space. Facing such problems, a direction vectors based co-evolutionary multi-objective optimization algorithm (DVCMOA) that introduces the decomposition idea from MOEA/D to co-evolutionary algorithms is proposed in this paper. It is novel in the sense that DVCMOA applies the concept of direction vectors to co-evolutionary algorithms. DVCMOA first divides the entire population into several subpopulations on the basis of the initial direction vectors in the objective space. Then, it solves MOPs through the co-evolutionary interaction among the subpopulations in which individuals are classified according to their direction vectors. Finally, it explores the less developed regions to maintain the relatively uniform distribution of the solution space. In this way, DVCMOA has advantages in convergence, diversity and uniform distribution of the non-dominated solution set, which are explained through comparison with other state-of-the-art multi-objective optimization evolutionary algorithms (MOEAs) in this paper. DVCMOA is shown to be effective on 6 multi-objective 0–1 knapsack problems.
    Information Sciences 04/2013; 228:90–112. · 3.64 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: The goal of this research is to study the incorporation of gradient-based information when designing Multi-objective Evolutionary Algorithms (MOEAs). We analyze the benefits, and challenges, of using these well developed mathematical programming techniques in order to get hybrid MOEAs. Since we expect the new hybrid algorithms to search effectively and more efficiently than currently available MOEAs, a deeper study of the balance between the computational and the benefits of this coupling is highly necessary.
    Genetic and Evolutionary Computation Conference, GECCO 2010, Proceedings, Portland, Oregon, USA, July 7-11, 2010, Companion Material; 01/2010