Pelikan, M., & Goldberg, D. E. (2001). Escaping hierarchical traps with competent genetic algorithms. Proceedings of the Genetic and Evolutionary Computation Conference (GECCO2001) , 511--518.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....find a factorization given a sample vector of data, we use an incremental greedy algorithm to minimize a metric that represents a trade o# between the likelihood and the complexity of the estimated probability distribution. This is a common approach that has been observed to give good results [2, 8, 12, 14]. The factorization learning algorithm starts from the univariate factorization in which all variables are independent of each other # = 0) 1) l 1) Each iteration, an operation that changes # is performed such that the value of the penalization metric decreases. This procedure ....

.... for selection when using a penalization metric, whereas otherwise non local structure models expressing similar dependencies would never have been regarded because of the large number of (redundant) parameters the impose [6] The use of local structures has been shown by Pelikan and Goldberg [14] to allow for e#cient optimization of very di#cult hierarchical deceptive optimization problems that exhibit dependencies between combinations of values for large groups of variables. To use default tables, the random keys for each factor # i are converted into integer permutations. The list of ....

M. Pelikan and D. E. Goldberg. Escaping hierarchical traps with competent genetic algorithms. In L. Spector, E. D. Goodman, A. Wu, W. B. Langdon, H.-M. Voigt, M. Gen, S. Sen, M. Dorigo, S. Pezeshk, M. H. Garzon, and E. Burke, editors, Proceedings of the GECCO--2001 Genetic and Evolutionary Computation Conference, pages 511--518. Morgan Kaufmann, 2001.

.... otherwise non local structure probability distributions expressing similar dependencies would never have been regarded because of the large number of (redundant) parameters they impose [9] Experimental verification of the usefulness of local structures has been given by Pelikan and Goldberg [16]. They show that using local structures in the iterated estimation of probability distributions for binary random variables allows for e#cient optimization of very di#cult hierarchical deceptive optimization problems that exhibit dependencies between combinations of values for large groups of ....

....random variables given a sample vector of data, we use an incremental greedy algorithm to minimize a metric that represents a trade o# between the likelihood and the complexity of the estimated probability distribution. This is a common approach that has been observed to give good results [4, 11, 13, 16]. Two of such commonly known metrics that have often proved to be successful, are known as the Akaike Information Criterion (AIC) and the Bayesian Information Criterion (BIC) Both metrics score a probability distribution by its negative log likelihood, but add a penalty term. This penalty term ....

[Article contains additional citation context not shown here]

M. Pelikan and D. E. Goldberg. Escaping hierarchical traps with competent genetic algorithms. In L. Spector, E. D. Goodman, A. Wu, W. B. Langdon, H.-M. Voigt, M. Gen, S. Sen, M. Dorigo, S. Pezeshk, M. H. Garzon, and E. Burke, editors, Proceedings of the GECCO-- 2001.

....learning is that the indegree of the Bayesian network can represent the gene relations of order k. Hence the population distribution is estimated with the information of building block groupings. The proposed algorithm is experimented on additively and hierarchically composed k deceptive problems [51, 52]. 17 PARALLEL GENETIC ALGORITHMS The main goal of this chapter is to summarize the recent research on parallel genetic algorithms (PGAs) In the past few years, PGAs have been used to solve di#cult problems. Hard problems need a bigger population and this translates directly into higher ....

....and the a#nity, it is possible to define some trap functions [22] where finding the linkage is important in the performance of the genetic algorithms. Also the linkage of a permutation could be observed in this way. The deceptive functions that are used in other linkage learning approaches like [30, 48, 38, 50, 51, 52] are not suitable for our purposes, since we aim to find the 55 f 3deceptive (u) 8 0.8 if u = 2 0 if u = 1 uis the number of ones in the 3 bit string 0.2 0.4 0.6 0.8 1 0 1 2 3 Figure 4.10: A trap function of order 3 and the corresponding graph. d f intdec (d) 0, # 1 ] ....

Martin Pelikan and David E. Goldberg. Escaping hierarchical traps with competent genetic algorithms. In L. Spector, E.D. Goodman, and A. Wu et.al., editors, Proceedings of the Genetic and Evolutionary Computation Conference (GECCO2001.

....hill climber [3] There remain some classes of problem, however, for which the existence of a population is desirable, if not essential. Examples include Watson Pollock s HIFF function [13] Holland s hyperplane defined functions [5] and Pelikan and Goldberg s hierarchical trap functions [10]. These are all problems that require a diverse population to be maintained in order for the optimal solution to be discovered. In biology, diversity is important for several reasons; it allows a population to adapt more rapidly to changing environments (e.g. in a host parasite situation [9] and ....

Pelikan, M. & Goldberg, D. E., Escaping Hierarchical Traps with Competent Genetic Algorithms, IlliGAL Technical Report

No context found.

Pelikan, M., & Goldberg, D. E. (2001). Escaping hierarchical traps with competent genetic algorithms. Proceedings of the Genetic and Evolutionary Computation Conference (GECCO2001) , 511--518.

No context found.

M. Pelikan and D. E. Goldberg. Escaping hierarchical traps with competent genetic algorithms. Proceedings of the Genetic and Evolutionary Computation Conference (GECCO2001.

No context found.

Pelikan, M., & Goldberg, D. E. (2001). Escaping hierarchical traps with competent genetic algorithms. Proceedings of the Genetic and Evolutionary Computation Conference (GECCO2001) , 511--518.

No context found.

Pelikan, M. and Goldberg, D. E., 2001, Escaping hierarchical traps with competent genetic algorithms, in: Proc. Genetic and Evolutionary Computation Conf., pp. 511--518.

No context found.

Pelikan, M., Goldberg, D.E.: Escaping hierarchical traps with competent genetic algorithms. In Spector et al., L., ed.: Proceedings of the Genetic and Evolutionary Computation Conference, GECCO-01, Morgan Kaufmann (2001) 511--518

No context found.

Martin Pelikan and David E. Goldberg. Escaping hierarchical traps with competent genetic algorithms. In L. Spector et al., editor, Proceedings of the Genetic and Evolutionary Computation Conference, GECCO-01, pages 511--518. Morgan Kaufmann, 2001.

No context found.

Martin Pelikan and David E. Goldberg. Escaping hierarchical traps with competent genetic algorithms. In L. Spector, E.D. Goodman, A. Wu, W. B. Langdon, H.- M. Voigt, M. Gen, S. Sen, M. Dorigo, S. Pezeshk, M. H. Garzon, and E. Burke, editors, Proceedings of the Genetic and Evolutionary Computation Conference, GECCO-2001, pages 511--518, 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