Paper Info
Title
A non-cooperative game for faster convergence in cooperative coevolution for multi-objective optimization
Abstract
Cooperative coevolution is an approach for evolving solutions from different populations which are evaluated based on how well they perform together. The advantage of cooperative coevolutionary algorithms is the decomposition of the problem which allows us to learn different parts of the problem instead of the whole problem at once. However, previous research within the field of global optimization has shown that cooperative coevolutionary algorithms are biased towards equilibrium states. Since studies concerning cooperative coevolutionary algorithms used for solving multi-objective optimization problems were initiated, no attention has been paid to this issue. In this paper, we show empirical evidence of the existence of these problems within the multi-objective optimization field and present a novel cooperative coevolution framework which, through the use of the concept of Nash equilibrium, alleviates some of those optimization-related pathologies present in cooperative coevolutionary algorithms. We compare our proposed algorithm with respect to two algorithms that make use of the cooperative coevolutionary model to multi-objective optimization, NSCCGA (that makes use of Potter's coevolutionary model) and GCEA (a game theory based coevo-lutionary algorithm). The computational effort required by each algorithm (measured in terms of the number of fitness function evaluations) is also analyzed. Our preliminary results indicate that the proposed framework clearly outperforms the results of the aforementioned algorithms when using the Deb-Thiele-Laumanns-Zitzler (DTLZ) and the Zitzler-Deb-Thiele (ZDT) test suites.
Year
DOI
Venue
2015
10.1109/CEC.2015.7256881
2015 IEEE Congress on Evolutionary Computation (CEC)
Keywords
Field
DocType
noncooperative game,convergence,multiobjective optimization,cooperative coevolutionary algorithms,global optimization,equilibrium states,Nash equilibrium,optimization-related pathologies,NSCCGA,Potter coevolutionary model,Deb-Thiele-Laumanns-Zitzler test suites,DTLZ test suites,Zitzler-Deb-Thiele test suites,ZDT test suites
Mathematical optimization,Global optimization,Computer science,Cooperative coevolution,Multi-objective optimization,Fitness function,Artificial intelligence,Game theory,Nash equilibrium,Non-cooperative game,Optimization problem,Machine learning
Conference
ISSN
Citations 
PageRank 
1089-778X
1
0.36
References 
Authors
9
2
Name
Order
Citations
PageRank
Luis Miguel Antonio1583.92
C. A. Coello Coello25799427.99