ABSTRACT
This paper introduces the Objective Fitness Correlation, a new tool to analyze the evaluation accuracy of coevolutionary algorithms. Accurate evaluation is an essential ingredient in creating adequate coevolutionary dynamics. Based on the notion of a solution concept, a new definition for objective fitness in coevolution is provided. The correlation between the objective fitness and the subjective fitness used in a coevolutionary algorithm yields the Objective Fitness Correlation. The OFC measure is applied to three coevolutionary evaluation methods. It is found that the Objective Fitness Correlation varies substantially over time. Moreover, a high OFC is found to correspond to periods where the algorithm is able to increase the objective quality of individuals. This is evidence of the utility of OFC as a measure to evaluate and compare coevolutionary evaluation mechanisms. The Objective Fitness Correlation (OFC) provides a precise analytical tool to measure the accuracy of evaluation in coevolutionary algorithms.
- Edwin D. De Jong and Jordan B. Pollack. Ideal evaluation from coevolution. Evolutionary Computation, 12(2):159--192, 2004. Google ScholarDigital Library
- Richard A. Watson and Jordan B. Pollack. Coevolutionary dynamics in a minimal substrate. In Spector, L., et al., editor, Proceedings of the Genetic and Evolutionary Computation Conference, GECCO-01, pages 702--709, San Francisco, CA, 2001. Morgan Kaufmann.Google Scholar
- Sevan G. Ficici. Solution Concepts in Coevolutionary Algorithms. PhD thesis, Brandeis University, 2004. Google ScholarDigital Library
- Sean Luke and RPaul Wiegand. Guaranteeing coevolutionary objective measures. In Kenneth~A. De Jong, Riccardo Poli, and Jonathan~E. Rowe, editors, Foundations of Genetic Algorithms 7, pages 237--252. Morgan Kaufmann, San Francisco, 2003.Google Scholar
- Susan L. Epstein. Toward an ideal trainer. Machine Learning, 15(3):251--277, 1994. Google ScholarDigital Library
- Hugues Juillé and Jordan B. Pollack. Coevolving the "ideal" trainer: Application to the discovery of cellular automata rules. In John R. Koza et al., editor, Proceedings of the Third Annual Genetic Programming Conference, pages 519--527, San Francisco, CA, USA, 1998. Morgan Kaufmann.Google Scholar
- Anthony Bucci and Jordan B. Pollack. Order-theoretic analysis of coevolution problems: Coevolutionary statics. In Proceedings of the GECCO-02 Workshop on Coevolution: Understanding Coevolution, 2002.Google Scholar
- E. Popovici and Jong K. Jong. Relationships between internal and external metrics in co-evolution. In David Corne et al., editor, Proceedings of the 2005 IEEE Congress on Evolutionary Computation, volume 3, pages 2800--2807, Edinburgh, Scotland, UK, 2-5 September 2005. IEEE Press.Google ScholarCross Ref
- D. Cliff and G. F. Miller. Tracking the Red Queen: Measurements of adaptive progress in co-evolutionary simulations. In FMorán, AMoreno, JJ. Merelo, and PChacín, editors, Proceedings of the Third European Conference on Artificial Life: Advances in Artificial Life, volume 929 of LNAI, pages 200--218, Berlin, 1995. Springer. Google ScholarDigital Library
- Sevan G. Ficici and Jordan~B. Pollack. Challenges in coevolutionary learning: Arms-race dynamics, open-endedness, and mediocre stable states. In CAdami, RBelew, HKitano, and TTaylor, editors, Proceedings of the Sixth International Conference on Artificial Life. The MIT Press, 1998. Google ScholarDigital Library
- Kenneth O. Stanley and Risto Miikkulainen. The dominance tournament method of monitoring progress in coevolution. In Alwyn M. Barry, editor, GECCO 2002: Proceedings of the Bird of a Feather Workshops, Genetic and Evolutionary Computation Conference, pages 242--248, New York, 8 July 2002. AAAI.Google Scholar
- Sevan G. Ficici and Jordan B. Pollack. A game-theoretic approach to the simple coevolutionary algorithm. In Schoenauer, M., et al., editor, Parallel Problem Solving from Nature, PPSN-VI, volume 1917 of LNCS, Berlin, 2000. Springer. Google ScholarDigital Library
- Richard A. Watson and Jordan~B. Pollack. Symbiotic combination as an alternative to sexual recombination in genetic algorithms. In Schoenauer, M., et al., editor, Parallel Problem Solving from Nature, PPSN-VI, volume 1917 of LNCS, pages 425--434, Berlin, 2000. Springer. Google ScholarDigital Library
- Ari Bader-Natal and Jordan~B. Pollack. Towards metrics and visualizations sensitive to coevolutionary failures. In Mitchell A. Potter and RPaul Wiegand, editors, Coevolutionary and Coadaptive Systems: Papers from the AAAI Fall Symposium, Technical Report FS-05-03, pages 1--8, Menlo Park, California, 2005. AAAI Press.Google Scholar
- Edwin D. De Jong. The MaxSolve algorithm for coevolution. In Hans-Georg Beyer, editor, Proceedings of the Genetic and Evolutionary Computation Conference, GECCO--05, pages 483--489. ACM Press, 2005. Google ScholarDigital Library
- Richard A. Watson. Personal communication, 2003.Google Scholar
- Edwin D. De Jong. Intransitivity in coevolution. In Xin Yao et al., editor, Parallel Problem Solving from Nature -- PPSN VIII, volume 3242 of LNCS, pages 843--851, Birmingham, UK, 18--22 September 2004. Springer-Verlag.Google Scholar
- Anthony Bucci, Jordan~B. Pollack, and Edwin D. De Jong. Automated extraction of problem structure. In Proceedings of the Genetic and Evolutionary Computation Conference, GECCO--04, pages 501--512, 2004.Google Scholar
- Sevan G. Ficici and Jordan B. Pollack. Pareto optimality in coevolutionary learning. In Jozef Kelemen, editor, Sixth European Conference on Artificial Life, pages 316--325, Berlin, 2001. Springer. Google ScholarDigital Library
- Christopher D. Rosin. Coevolutionary Search among Adversaries. PhD thesis, University of California, San Diego, CA, 1997. Google ScholarDigital Library
Recommendations
Correlation analysis of coupled fitness landscapes: Research Articles
The correlation structure of fitness landscapes is a much used measure to characterize and classify various types of landscapes. However, analyzing the correlation structure of fitness landscapes has so far been restricted to static landscapes only. ...
A cooperative coevolutionary biogeography-based optimizer
With its unique migration operator and mutation operator, Biogeography-Based Optimization (BBO), which simulates migration of species in natural biogeography, is different from existing evolutionary algorithms, but it has shortcomings such as poor ...
Solution and Fitness Evolution (SAFE): Coevolving Solutions and Their Objective Functions
Genetic ProgrammingAbstractWe recently highlighted a fundamental problem recognized to confound algorithmic optimization, namely, conflating the objective with the objective function. Even when the former is well defined, the latter may not be obvious, e.g., in learning a ...
Comments