D. B. Fogel and A. Ghozeil. Using fitness distributions to design more efficient evolutionary computations. In T. Fukuda, editor, Proceedings of the Third IEEE International Conference on Evolutionary Computation, pages 11--19. IEEE, 1996.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....surface has multiple domains of attraction. Grefenstette [7] used the probability distribution of offspring fitnesses indexed by the mean parents fitnesses to yield a first order linear approximation of the suitability of a genetic operator. Another approach was suggested by Fogel and Ghozeil [5]: to compare different selection schemes and operators, they use the mean improvement of offspring, and concludes that the uniform intermediate recombinations yields better results (higher mean improvement) than the one point crossover, for all three selection methods they tested, on some ....

....fitness can be better (in mean) than its parents : the correlation can be maximal when no improvement is to be expected However unrealistic this particular case can be, it nevertheless shows the limits of M FOC as a measure of operator s efficiency. In that line, as noted by Fogel and Ghozeil [5], M FOC cannot be used on problems that yield zero mean difference between parents and offspring fitnesses, as in the case of linear fitness functions with real valued representations. The correlation is always maximal, and cannot reflect the high dependency of their convergence rate on the the ....

[Article contains additional citation context not shown here]

D. B. Fogel and A. Ghozeil. Using fitness distributions to design more efficient evolutionary computations. In T. Fukuda, editor, Proceedings of the Third IEEE International Conference on Evolutionary Computation, pages 11--19. IEEE, 1996.

....the PCR reaction) DE is a technology with immense potential for novel and highly beneficial protein products, but the search spaces are massive, and much depends on appropriate design and parameterization of DE strategies. 1. 3 A Note on Related Work Seminal work by Fogel and co authors (e.g. [8] and subsequent papers) relates closely to that described here, but there are subtle differences in approach and applicability which are worth noting. In [8] and related work, Fogel et al. like us, essentially recognize that repeated applications of genetic operators yield valuable information ....

....depends on appropriate design and parameterization of DE strategies. 1.3 A Note on Related Work Seminal work by Fogel and co authors (e.g. 8] and subsequent papers) relates closely to that described here, but there are subtle differences in approach and applicability which are worth noting. In [8] and related work, Fogel et al. like us, essentially recognize that repeated applications of genetic operators yield valuable information which can be employed to choose and parameterize the operator(s) more wisely. In [8] however, this is done essentially using extensive offiine prior sampling ....

[Article contains additional citation context not shown here]

Fogel, D.B. and Ghozeil, A. (1996). Using Fitness Distributions to Design More Efficient Evolutionary Computations, in Proceedings of the 3 rd International Conference on Evolu- tionary Computation, IEEE, pp. 11-19.

....absolute benefit. An efficient method to calculate the partial derivatives with respect to the numerical coefficients of the evolved expressions is presented. 2 Fitness distributions Fitness distribution (FD) analysis has been proposed as a tool for designing efficient evolutionary algorithms [9, 13, 14]. The FD of a variation operator is described as the distribution of the offspring fitness given the fitness of the parents [13] This distribution is generally quite complex and difficult to compute, so the analysis is restricted to certain features of the distribution like the probability of ....

....But it is not only important how often offspring are better than their parents, but also how much better they are [1] This can be expressed by the the expected improvement EI # = E[#(#(a) #(a) where E[ denotes the expectation. The benefit B# [27] combines both measures (see also [9]) B# = E[max 0, #(#(a) #(a) 1) None of these measures necessarily reflects evolvability which is defined as the ability of an algorithm to produce individuals fitter than any existing, one of the key properties determining the quality of evolutionary search [1] If we replace the ....

D. B. Fogel and A. Ghozeil. Using fitness distributions to design more efficient evolutionary computations. In Proc. 1996 IEEE Int. Conf. on Evolutionary Computation, pp. 11-- 19. IEEE Press, 1996.

.... method handle the optimization of the parameters of evolution As far as we know, the only possibility studied so far was that of the 1 5 rule [24] developped by careful analysis of two simplifed models (the sphere and the corridor) However, its generalization to other situations is not clear [8]. Moreover, this deterministic method is limited in that it can only globally handle the vector of standard deviations (oe i ) since it simply counts the number of successes and failures (offspring more or less fit than the parents) counting does not supply enough information to increase the ....

D. B. Fogel and A. Ghozeil. Using fitness distributions to design more efficient evolutionary computations. In T. Fukuda, editor, Proceedings of the Third IEEE International Conference on Evolutionary Computation, pages 11--19. IEEE, 1996.

....on random sampling of the search space. We cite from the field of evolutionary algorithms: Fitness Distance relations, first proposed in [7] and successfully used to choose problem dependent random initialization procedures [10, 13] Fitness Improvement of evolution operators, first proposed in [4], then extended and successfully used to choose binary crossover operators [11] and representations [8] However, even if such heuristics can guide the a priori choice of some EA parameters, they do not give significant information about landscape structure, for instance, recent work suggests that ....

D. B. Fogel and A. Ghozeil, Using fitness distributions to design more efficient evolutionary computations, in Proceedings of the Third IEEE International Conference on Evolutionary Computation, T. Fukuda, ed., IEEE, 1996, pages 11-19.

No context found.

Fogel, D.B., Ghozeil, A.: Using fitness distributions to design more efficient evolutionary computations. Proc. of 1996 IEEE Conf. on Evol. Comp., Keynote Lecture, IEEE Press, NY (1996) 11-19.

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