Richard A. Watson and Jordan B. Pollack. Coevolutionary dynamics in a minimal substrate. In L. Spector, E. Goodman, A. Wu, W.B. Langdon, H.-M. Voigt, M. Gen, S. Sen, M. Dorigo, S. Pezeshk, M. Garzon, and E. Burke, editors, Proceedings of the Genetic and Evolutionary Computation Conference, GECCO-2001, pages 702--709, San Francisco, CA, 2001. Morgan Kaufmann.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....algorithms has focused on their complicated dynamics. For example, considerable effort has been spent trying to understand how one can measure progress in a system where individual fitnesses are subjective in order to help identify some of the pathological behaviors exhibited by these algorithms [5 7]. Additionally, some basic theoretical work uses ideas from simple genetic algorithm theory provided by [8] and applies them to competitive coevolution [9] That work explores the mechanics of a simple competitive coevolutionary algorithm from an evolutionary game theoretic viewpoint. 10] ....

R. Watson and J. Pollack. Coevolutionary dynamics in a minimal substrate. In Spector [15], pages 702--709.

....from the relative nature of fitness assessment individuals receive a fitness based upon their success against contemporary opponents. In general, coevolutionary systems are difficult to direct. Individuals may overfit their contemporary competitors, resulting in potentially brittle solutions [5, 6]. Rather than enter a progressive arms race, competing populations may stabilise into a suboptimal equilibrium, or mediocre stable state [3, 7, 8, 6] As individuals are only rewarded for out performing their contemporary oponents, it is possible for earlier adaptations to be lost, potentially ....

.... contemporary oponents, it is possible for earlier adaptations to be lost, potentially leading to cycling [9, 10, 7, 8, 6] Finally, if one population outperforms the other to the extent that every opponent is beaten, the gradient for selection disappears and the populations disengage and drift [5, 6, 11]. As drift during periods of disengagement is random rather than neutral, near optimal populations are likely to degenerate. Although there are methods for counter acting particular coevolutionary problems, e.g. fitness sharing and the hall of fame [12] few of these address the problem of ....

[Article contains additional citation context not shown here]

Watson, R.A., Pollack, J.B.: Coevolutionary dynamics in a minimal substrate. In: Proceedings of the Genetic and Evolutionary Computation Conference. (2001) 702--709

....Graham Kendall for his invitation to submit the paper, Aaron Davidson for discussions on the work of the Alberta group, and to Josh Knowles for suggestions on diversity maintenance. well, they provide not only varied opposition, but also opposition that is graded to an appropriate diculty level [2]: as evolution improves one strategy, it also improves the competition (the familiar arms race scenario) However, coevolutionary algorithms do not always work as intended, and there are a number of pitfalls in their use [2] One such pitfall involves intransitive superiority relationships. ....

.... but also opposition that is graded to an appropriate diculty level [2] as evolution improves one strategy, it also improves the competition (the familiar arms race scenario) However, coevolutionary algorithms do not always work as intended, and there are a number of pitfalls in their use [2]. One such pitfall involves intransitive superiority relationships. That is, although some player A might be beaten by some other player B, and B may in turn be beaten by C, it may not be the case that C beats A. The existence of intransitive superiority relationships can mean that although a ....

R. A. Watson and J. B. Pollack, \Coevolutionary dynamics in a minimal substrate," in Proceedings of the Genetic and Evolutionary Computation Conference, GECCO-2001.

....of Leeds, Leeds, LS2 9JT Tel: 44 (0)113 233 5322, Fax: 44 (0)113 233 5468, Email: fjohnc,sethg comp.leeds.ac. uk Abstract Inspired by the virulence of natural parasites, a novel approach is developed to tackle disengagement, a detrimental phenomenon coevolutionary systems sometimes experience [1]. After demonstrating bene cial results in a simple model, minimumcomparison sorting networks are coevolved, with results suggesting that moderating parasite virulence can help in practical problem domains. I. INTRODUCTION Arti cial coevolution can be divided into two main categories, ....

....to analyse it is hard to tell whether sustained competition is leading to smooth directional evolutionary progress, to noisy ts and starts, or to endless cycling through di erent evolutionary unstable strategies. In this paper, however, we concentrate on the related problem of disengagement [1]. As coevolutionary algorithms assess individuals through pitting them against opponents, there arises the problem of ensuring that opponents are worth beating. Disengagement occurs when a population faces opponents that are either all too easy or too dicult to beat. In such a situation all ....

[Article contains additional citation context not shown here]

R. A. Watson and J. B. Pollack. Coevolutionary dynamics in a minimal substrate. In Spencer et al. [16], pages 702-709. E.g., pressure on parasites to remain only moderately virulent, may be responsible for the failure of hosts to achieve perfect solutions in the counting-ones problem.

.... success and efficiency of co evolutionary search algorithms, are also due to other reasons [20] typically cited as the possibility of an adaptive and measurable performance gradient (i.e. the gain in efficiency of the evaluation of evolving solutions [11] which could otherwise be hard to engineer [30], focusing adaptation on any relevant target, i.e. on those aspects of a task that have not yet been optimised [11,13,30] the possible automatic adjustment of the selection gradient which is imposed on the evolving solutions [15] and finally on the above cited potential open ended nature of ....

.... the possibility of an adaptive and measurable performance gradient (i.e. the gain in efficiency of the evaluation of evolving solutions [11] which could otherwise be hard to engineer [30] focusing adaptation on any relevant target, i.e. on those aspects of a task that have not yet been optimised [11,13,30], the possible automatic adjustment of the selection gradient which is imposed on the evolving solutions [15] and finally on the above cited potential open ended nature of these systems [24,8] Due to these reasons, the idea of reciprocal and iterated adaptation has become increasingly popular in ....

Richard A. Watson, Jordan B. Pollack (2001), "Coevolutionary Dynamics in a Minimal Substrate". In L. Spector et al. (Eds.), GECCO-01 -- Procs. of the Genetic and Evolutionary Computation Conf., Morgan Kaufmann, pp. 702-709.

No context found.

Richard A. Watson and Jordan B. Pollack. Coevolutionary dynamics in a minimal substrate. In L. Spector, E. Goodman, A. Wu, W.B. Langdon, H.-M. Voigt, M. Gen, S. Sen, M. Dorigo, S. Pezeshk, M. Garzon, and E. Burke, editors, Proceedings of the Genetic and Evolutionary Computation Conference, GECCO-2001, pages 702--709, San Francisco, CA, 2001. Morgan Kaufmann.

No context found.

Watson, R. A., and Pollack, J. B. 2001. Coevolutionary dynamics in a minimal substrate. In Proceedings of the Genetic and Evolutionary Computation Conference (GECCO 2001), 702--709.

No context found.

Watson, R., Pollack, J.B.: Coevolutionary dynamics in a minimal substrate. In L. Spector et al., ed.: Proceedings of the Genetic and Evolutionary Computation Conference, GECCO2001, San Francisco, CA, Morgan Kaufmann Publishers (2001)

No context found.

Watson, R. A., and Pollack, J. B. 2001. Coevolutionary dynamics in a minimal substrate. In Spector, L., et al., eds., Proceedings of the 2001 Genetic and Evolutionary Computation Conference. Morgan Kaufmann.

No context found.

Watson, R.A., Pollack, J.B.: Coevolutionary dynamics in a minimal substrate. In Lee Spector et. al., ed.: Proceedings of the 2001 Genetic and Evolutionary Computation Conference, Morgan Kaufmann (2001)

No context found.

R. A. Watson and J. B. Pollack. Coevolutionary dynamics in a minimal substrate. In Lee Spector, Erik D. Goodman, Annie Wu, W. B. Langdon, Hans-Michael Voigt, Mitsuo Gen, Sandip Sen, Marco Dorigo, Shahram Pezeshk, Max H. Garzon, , and Edmund Burke, editors, Proceedings of the 2001 Genetic and Evolutionary Computation Conference. Morgan Kaufmann, 2001.

No context found.

R. A. Watson and J. B. Pollack. Coevolutionary dynamics in a minimal substrate. In L. Spector et al., editors, Proc. 2001 Genetic and Evolutionary Computation Conf., pages 702--709. Morgan Kaufmann, 2001.

No context found.

Richard A. Watson and Jordan B. Pollack. Coevolutionary dynamics in a minimal substrate. In L. Spector, E. Goodman, A. Wu, W.B. Langdon, H.-M. Voigt, M. Gen, S. Sen, M. Dorigo, S. Pezeshk, M. Garzon, and E. Burke, editors, Proceedings of the Genetic and Evolutionary Computation Conference, GECCO-2001, pages 702--709, San Francisco, CA, 2001. Morgan Kaufmann.

No context found.

Watson, R., Pollack, J.B.: Coevolutionary dynamics in a minimal substrate. In L. Spector et al., ed.: Proceedings of the Genetic and Evolutionary Computation Conference, GECCO2001, San Francisco, CA, Morgan Kaufmann Publishers (2001)

....systems, rendering it difficult or impossible to tell if the algorithm has accomplished something useful. Mediocre stable states [5] arise when collusion permits suboptimal individuals to appear better than they are. Intransitive superiority cycles, and the related problem of overspecialization [14], can cause coevolutionary dynamics to cycle through a set of suboptimal individuals without making progress. These issues are apparently related. Indeed, it is unclear how one can even reasonably discuss notions like goodness or progress in a coevolutionary setting. Understanding these issues ....

....finite set of players a i 2 S such that p(a i 1 ; a i ) 1 and p(a ; a i 1 ) 0 for 1 i n 1, but p(a n ; a 0 ) 1. Transitivity would dictate p(a 0 ; a n ) 1. Coevolutionary dynamics operating on such a game can become stuck cycling amongst the a i without ever making real progress. See [14] for a discussion of this issue. One of the promises of Pareto coevolution is that it can help with intransitive cycles by revealing the true relationship among the individuals in a cycle. We will show there is a close relationship between the transitivity of p and the Pareto dominance ....

Richard Watson and Jordan Pollack. Coevolutionary dynamics in a minimal substrate. In L. Spector, E. Goodman, A. Wu, W.B. Langdon, H.-M. Voigt, M. Gen, S. Sen, M. Dorigo, S. Pezeshk, M. Garzon, and E. Burke, editors, Proceedings of the Genetic and Evolutionary Computation Conference, GECCO-2001.

....progress, we must make observations which allow us to see if the algorithm really is climbing S lt:p . Even better, we should arrange our algorithms to guarantee, as much as possible, progress up the order S lt:p . Pareto coevolution involves heuristics for achieving this goal. Finally, focusing [15] refers to the ability of coevolving opponents to challenge one another by testing weak dimensions of performance. An issue which arises in this context is overspecialization. In the language of our framework, T is well focused when the tests in T are informative, meaning they indicate many ....

Richard Watson and Jordan Pollack. Coevolutionary dynamics in a minimal substrate. In L. Spector, E. Goodman, A. Wu, W.B. Langdon, H.-M. Voigt, M. Gen, S. Sen, M. Dorigo, S. Pezeshk, M. Garzon, and E. Burke, editors, Proceedings of the Genetic and Evolutionary Computation Conference, GECCO2001.

....progress, we must make observations which allow us to see if the algorithm really is climbing S t:p . Even better, we should arrange our algorithms to guarantee, as much as possible, progress up the order S t:p . Pareto coevolution involves heuristics for achieving this goal. Finally, focusing [15] refers to the ability of coevolving opponents to challenge one another by testing weak dimensions of performance. An issue which arises in this context is overspecialization. In the language of our framework, T is well focused when the tests in are informative, meaning they indicate many ....

Richard Watson and Jordan Pollack. Coevolutionary dynamics in a minimal substrate. In L. Spector, E. Goodman, A. Wu, W.B. Langdon, H.-M. Voigt, M. Gen, S. Sen, M. Dorigo, S. Pezeshk, M. Garzon, and E. Burke, editors, Proceedings of the Genetic and Evolutionary Computation Conference, GECCO-2001.

No context found.

R. A. Watson and J. B. Pollack. Coevolutionary dynamics in a minimal substrate. In Proceedings of the Genetic and Evolutionary Computation Conference, GECCO-01, pages 702--709, San Francisco, CA, 2001. Morgan Kaufmann.

No context found.

R. A. Watson and J. B. Pollack. Coevolutionary dynamics in a minimal substrate. In L. Spector et al., editors, Proc. of the 2001.

No context found.

R. A. Watson and J. B. Pollack. Coevolutionary dynamics in a minimal substrate. In L. Spector, E. D. Goodman, A. Wu, and W. B. Langdon, editors, Procs GECCO 2001. Morgan Kaufmann, 2001.

No context found.

Richard A. Watson and Jordan B. Pollack. Coevolutionary dynamics in a minimal substrate. In L. Spector, E. Goodman, A. Wu, W.B. Langdon, H.-M. Voigt, M. Gen, S. Sen, M. Dorigo, S. Pezeshk, M. Garzon, and E. Burke, editors, Proceedings of the Genetic and Evolutionary Computation Conference, GECCO-01, pages 702--709, San Francisco, CA, 2001. Morgan Kaufmann.

No context found.

Watson, R.A., Pollack, J.B.: Coevolutionary dynamics in a minimal substrate. In Spector, L., Goodman, E., Wu, A., Langdon, W., Voigt, H.M., Gen, M., Sen, S., Dorigo, M., Pezeshk, S., Garzon, M., Burke, E., eds.: Proceedings of the Genetic and Evolutionary Computation Conference, GECCO-01, San Francisco, CA, Morgan Kaufmann (2001) 702--709

No context found.

Richard A. Watson and Jordan B. Pollack. Coevolutionary dynamics in a minimal substrate. In L. Spector, E. Goodman, A. Wu, W.B. Langdon, H.-M. Voigt, M. Gen, S. Sen, M. Dorigo, S. Pezeshk, M. Garzon, and E. Burke, editors, Proceedings of the Genetic and Evolutionary Computation Conference, GECCO-01, pages 702--709, San Francisco, CA, 2001. Morgan Kaufmann.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC