H. Muhlenbein, M. Gorges-Schleuter, and O. Kramer, \Evolution Algorithms in Combinatorial Optimization," Parallel Computing, 7, (1988), 65-88.  Home/Search   Document Not in Database   Summary   Related Articles   Check

.... Syswerda (1991) 17] Position Based Crossover (POS) Syswerda (1991) 17] Heuristic Crossover (HX) Grefenstette (1987) 5] Edge Recombination Crossover (ER) Whitley et al. 1989) 18] Sorted Match Crossover (SMX) Brady (1985) 1] Maximal Preservative Crossover (MPX) M uhlenbein et al. 1988) [13] Voting Recombination Crossover (VR) M uhlenbein (1989) 14] Alternating Position Crossover (AP) Larranaga et al. 1996) 9] Among these PMX, ER and POS are quoted to be the fastest operators as far as the number of necessary iterations to reach convergence is concerned [10, 16] The ....

Muhlenbein, H., M. Gorges-Schleuter, and O. Kramer. \Evolution Algorithms in Combinatorial Optimization." Parallel Computing, 7, 1988, pp. 65-85.

.... In addition to this bioinformatical application, common intervals also relate to the consecutive arrangement problem [2, 7, 8] and to cross over operators for genetic algorithms solving sequencing problems such as the traveling salesman problem or the single machine scheduling problem [3, 15, 18]. Recently, Uno and Yagiura [26] presented an optimal O(n K) time and O(n) space algorithm for nding all K n 2 common intervals of two permutations 1 and 2 of n elements. We generalized this algorithm to a family = 1 ; k ) of k 2 permutations in optimal O(kn K) time ....

H. Muhlenbein, M. Gorges-Schleuter, and O. Kramer. Evolution algorithms in combinatorial optimization. Parallel Comput., 7:65-85, 1988.

....researchers (see, e.g. 9] and the references therein) have considered examining the correlation between tness and distances between points on discrete tness landscapes. These studies include the cost versus average distance of an optimum point to all other points within a sample of local optima [2, 8] and cost versus distance of local optima from the best optima found [6] These studies on arti cial and combinatorial optimization landscapes have indicated that a Massif Central [6] or big valley [2] structure seems to exist in many landscapes. That is, perhaps not surprisingly, cost in ....

M uhlenbein, H., Gorges-Schleuter, M., and Kr amer, O. Evolution algorithms in combinatorial optimization. Parallel Computing 7 (1988), 65-85.

....algorithm (abbreviated as GA; also called as evolutionary computation 1 ) 34, 68, 76, 94, 115, 149] simulated annealing (abbreviated as SA) 1, 3, 24, 86, 92] tabu search (abbreviated as TS) 55, 59, 62, 64, 75] and so on. Among variants of these are genetic local search (abbreviated as GLS) [17, 18, 49, 82, 95, 114, 146, 151], which incorporates LS into GA, greedy randomized adaptive search procedure (abbreviated as GRASP) 43, 45, 46, 47, 97, 99, 137] which uses randomized greedy methods to generate initial solutions for LS, iterated local search (abbreviated as ILS) 85, 110] which uses good solutions found in the ....

....them. As the word genetic algorithm (GA) is also used to mean the general framework including GLS, we use simple GA to denote the genetic algorithms which do not incorporate LS, if we want to distinguish them from GLS. The basic idea of GLS was proposed in [18] Early references such as [34, 68, 82, 83, 111, 113, 114, 146, 151] have also mentioned the idea of GLS. To our knowledge, the word genetic local search rst appeared in [151] A similar but more general mechanism of generating initial solutions from many reference solutions is also proposed and is called the scatter search [53, 56, 57] While good solutions ....

[Article contains additional citation context not shown here]

H. Muhlenbein, M. Gorges-Schleuter and O. Kramer, \Evolution algorithms in combinatorial optimization," Parallel Computing, vol.7, pp.65-85, 1988.

....in the centre of a neighbourhood. Elitist replacement can not be transferred to local survival selection since global knowledge about the whole population would be necessary. 7 5. 3 Implementations of Local Population Models The first parallel implementation of a local model was done by M UHLENBEIN, GORGESSCHLEUTER und KR AMER (1988) on a shared memory multiprocessor ENCORE. Figure 4 summarizes some important implementations of the local model. Reference G N Local Mating Selection Computer (PEs) Neighbourhood Topology Local Survival Selection M UHLENBEIN et al. 1988) 16 5 rank proportional ENCORE (16) Ladder plus global ....

....of a local model was done by M UHLENBEIN, GORGESSCHLEUTER und KR AMER (1988) on a shared memory multiprocessor ENCORE. Figure 4 summarizes some important implementations of the local model. Reference G N Local Mating Selection Computer (PEs) Neighbourhood Topology Local Survival Selection M UHLENBEIN et al. 1988) 16 5 rank proportional ENCORE (16) Ladder plus global best GORGES SCHLEUTER (1989) 16,32,64 5,8 fitness rank prop. Parsitec T800 (64) Ladder, toroidal grid generational MANDERICK SPIESSENS (1989) 49 8 fitness prop. random sequential (1) toroidal X grid generational SPIESSENS ....

M UHLENBEIN, H., GORGES-SCHLEUTER, M., & KR AMER, O. 1988. Evolution Algorithms in Combinatorial Optimization. Parallel Computing, 7, 65--85.

....(UK) 360] Neural Computation, 467] Neural Networks, 338] Neural Parallel Sci. Comput, 63] Neural Parallel Sci. Comput. USA) 66] New Generation Computing Journal, 18] Nuclear Instruments Methods in Physics Research A, 38] Nuclear Technology, 317] Parallel Computing, [65, 210, 67, 74, 75, 442, 76, 449, 77] Parallel Computing (Netherlands) 531] Parallel Processing Letters, 384] PARS Mitteilungen, 143] Pattern Recognition, 81, 332] Physica D, 12, 32, 37] Physical Review C, Nuclear Physics, 336, 337] Proc. Inst. Mech. Eng. Part I: J. Syst. Control Eng. 162] Scienti c Computing in ....

....212, 282] Gitler, Daniel, 211] Giusti, Giuliano, 86, 199] Godart, C. 176] Godza, G. 283] Gogonea, Valentin, 236] Gold, S onke Sonnich, 488] Goldberg, David E. 210, 312] Goodman, Erik D. 88, 188, 189, 242, 341, 483] Gordon, Vahl Scott, 135, 148, 520] Gorges Schleuter, Martina, [93, 405, 406, 407, 408, 409, 410, 404, 442, 76] Grabski, W. 122] Green, David G. 22] Green, P. R. 116, 145] Greenwood, Garrison W. 136] Grefenstette, John J. 468] Grocholewska Czurylo, A. 28] Gruau, Fr ed eric C. 411] Gucht, Dirk Van, 58] Guidec, Fr ed eric, 237, 275, 72] Gupta, Ajay, 136] Hajela, Prabhat, 206, 286] ....

[Article contains additional citation context not shown here]

Heinz Muhlenbein, Martina Gorges-Schleuter, and O. Kramer. Evolution algorithms in combinatorial optimization. Parallel Computing, 7:65-85, April 1988. ga:Muhlenbein88a.

....would avoid doing a mistake more than once and would prefer making advantageous decisions again. For optimization a class of today s most popular heuristic approaches is known as genetic algorithms. Due to its widespread use and the vast amount of literature dealing with genetic algorithms, e.g. [2, 9, 16, 25, 28, 29, 33], a comprehensive review of research activities is doomed to failure. Thus, we stick to an outline of the fundamental ideas. The adjective genetic reveals the roots of these algorithms. Adapting the evolution strategy from natural life forms, the basic idea is to start with a set of (feasible) ....

M uhlenbein, H., Gorges--Schleuter, M., Kr amer, O., (1988), Evolution Algorithms in Combinatorial Optimization, Parallel Computing, Vol. 7, pp. 65--85

....an individual j represents a certain partition, determined by a map j . The i th component of a string is associated with the i th element of U containing the mapping goal which is an element of V . Several authors applied genetic algorithms to the graph partitioning problem, for instance [MGSK88] However, we will involve this approach into the above described hierarchical strategy. Here we focus onto our special 2 level scheme (3.4) Then genetic algorithms may be used as one of the parallel working algorithms in each hierarchical level. Yet, because of the large number of cones in Co ....

H. MÄuhlenbein, M. Gorges-Schleuter, and O. KrÄamer. Evolution Algorithm in Combinatorial Optimization. Parallel Computing, (7):65-88, 1988.

....parent 2 : 0 1 0 1 0 0 1 0 1 1 0 0 0 0 1; o spring: 0 0 0 1 1 1 0 0 1 1 0 0 1 0 0: The edge (5,6) occurred in both parents. However, it was not passed on to the o spring. 4.4. 6 Sorted Match Crossover The sorted match crossover operator was proposed by Brady [7] It (see also M uhlenbein et al. [40]) searches for subtours in both the parent tours which have the same length, which start in the same city, which end in the same city and which contain the same set of cities. If such subtours are found the cost of these substrings are determined. The o spring is constructed from the parent which ....

....6 5 7) These subtours have the same length, both begin in city 4, both end in city 7, and both contain the same cities. Suppose that the cost of the subtour (4 5 6 7) is higher than the cost of the subtour (4 6 5 7) Then, the following o spring is created: 1 2 3 4 6 5 7 8) M uhlenbein et al. [40] concluded that the sorted match crossover was useful in reducing the computation time, but that it is a weak scheme for crossover. 4.4.7 Maximal Preservative Crossover (MPX) The maximal preservative operator was introduced by M uhlenbein et al. 40] It works in a similar way to the PMX operator. ....

[Article contains additional citation context not shown here]

M uhlenbein, H., Gorges-Schleuter, M. and Kr amer, O. (1988), Evolution Algorithms in Combinatorial Optimization, Parallel Computing, 7, pp. 65-85.

....some of these methods in order to arrive at high quality solutions, particularly for large problem instances. For example, local search has been used in [7] 8] 18] 33] 34] to improve genetic algorithms (GAs) for the TSP. As a consequence, Gorges Schleuter and M uhlenbein [15] 16] [27] have proposed a GA where all individuals of the population are local minima with respect to the embedded local search method. Ulder et al. 35] compared this genetic local search (GLS) approach with other heuristics and observed that GLS is superior to simulated annealing as well as multi start ....

H. M¨uhlenbein, M. Gorges-Schleuter, and O. Kr¨amer, "Evolution Algorithms in Combinatorial Optimization," Parallel Computing, vol. 7, pp. 65--88, 1988.

....on a cleverly chosen energy function. It has recently been demonstrated that there is a strong correspondance between this elastic net al..gorithm and the Potts approach [9] 10] 11] Parallel to these developments genetic algorithms have been developed for solving these kind of problems [12][13] with extremely high quality results. Given the above mentioned scepticism towards the neural network approach and the relatively unknown success of the genetic approach we found it worthwhile to test these three different parallel distributed approaches on a common set of problems and compare the ....

....consist of two binary variables, which in the mean field theory treatment becomes two analog variables; N cities requires N 2 analog variables. In the elastic net case N cities only requires 2M(M N) analog variables; it is a more ecomical way of representing the problem. The Genetic Algorithm [13] [16] For details we refer the reader to refs. 13] 16] Here we briefly list the main steps and the parameters used. 1. Give the problem to M individuals. 2. Let each individual compute a local minimum (2 quick 2 ) 3. Let each individual choose partner for mating. In contrast to earlier ....

[Article contains additional citation context not shown here]

H. M¨uhlenbein, M. Gorges-Schleuter and O. Kr¨amer, "Evolution Algorithms in Combinatorial Optimization", Parallel Computing 7, 65 (1988);

....for food and produces offspring. In evolutionary algorithms, F (x i ) is called the fitness of individual i, x t 1 i is an offspring of x t i , and G is called the selection schedule. 3 Evolutionary algorithms A previous survey of search strategies based on evolution has been done in [M uhlenbein, Gorges Schleuter Kr amer, 1988] . We recall only the most important ones. Evolutionary algorithms which are driven mainly by mutation have been developed by Rechenberg [1973] and Schwefel [1981] for continuous parameter optimization. Their algorithms are called evolution strategies. Evolution Strategies STEP1: Create an ....

M¨uhlenbein, H., Gorges-Schleuter, M.& Kr¨amer, O. (1988). Evolution algorithms in combinatorial optimization. Parallel Computing, 7:65--88.

....]g: 1) The length of a common interval ( x A ; y A ] x B ; y B ] is de ned to be y A xA 1. Some genetic algorithms based on common intervals have been proposed for sequencing problems (e.g. traveling salesman problem, single machine scheduling problem, etc. and have exhibited good prospect [1, 2, 3, 4]. In this paper, we consider enumeration of all common intervals of length 2 to n. Three algorithms are proposed, which are improved versions of a simple O(n 2 ) time algorithm proposed in [5] 1. A simple O(n 2 ) time algorithm (called LHP) whose expected running time becomes O(n) for two ....

H. Muhlenbein, M. Gorges-Schleuter and O. Kramer, Evolution Algorithms in Combinatorial Optimization, Parallel Computing, 7 (1988), 65-85.

....indicated that pure GA approaches are outperformed by the conventional heuristics developed for the TSP, particularly when the problem size increases. The desire to improve the performance of GAs in TSP applications has motivated several researchers to build additional heuristic elements into GAs [21, 27]. The main possibilities to augment GAs with such heuristics are as follows: Instead of generating random tours which form the individuals of the initial population of a GA, it is possible to use a tour construction heuristic, such as the nearest neighbor (NN) heuristic or various insertion ....

....Edge NN incorporates greedy choices into the recombination step and additionally improves individuals by 2 and 3 changes. The tour lengths that have been achieved by edge NN for the 532 cities problem are 27949 (0.95 ) as the shortest tour and 28255 (2.06 ) as the average value. M uhlenbein [19, 21] has developed a parallel GA which is quite different to the ones described above. In this GA, each individual can be seen as an active entity which chooses itself a partner for mating and improves during its lifetime (i.e. local hill climbing is performed) A new crossover operator, called MPX ....

H. M¨uhlenbein, M. Gorges-Schleuter, and O. Kr¨amer, "Evolution Algorithms in Combinatorial Optimization," Parallel Computing, Vol. 7, pp. 65--88, 1988.

....heuristics with a population based strategy. Due to its intrinsic parallelism and the inherent asynchronicity of the method it is specially appealing for MIMD message passing parallel computers, such as those constructed from transputers. The approach is similar to that used by M uhlenbein [14] [15] [16] Brown et al. 1] Gorges Schleuter [3] and work performed by the Dynamics of Computation Group at Xerox PARC [4] We consider them as prototype examples of memetic algorithms in the sense described in Ref. 12] see also Ref. 5] A preliminary description of our work can also be found ....

H. M¨uhlenbein, "Evolution Algorithms in Combinatorial Optimization", Parallel Computing, 7, pp. 65, (1988).

....a traditional EA requires information about the fitnesses of all individuals during the reproduction phase. This kind of global knowledge makes an algorithm unsuitable for an efficient parallel implementation. Therefore, most parallel implementations of EAs base on local reproduction rules [7, 5, 13, 8, 12, 9, 14] which can be applied simultaneously to smaller subsets of the population. In order to be comparable to a standard GA, in [11] a localized proportionate selection was defined for a ring topology. In the following a more general definition is given which does not even depend on homogeneous ....

H. M¨uhlenbein, M. Gorges-Schleuter, and O. Kr¨amer. Evolution algorithms in combinatorial optimization. Parallel Computing, 7:65--88, 1988.

....by a procedure that repairs or projects the newly created vector in the space of feasible solutions. These particularities can also be seen as generalizations of the basic GA procedure. In fact, such generalizations have later been proposed and exploited by various authors (see for example [39, 42]) Let us quote: ffl departure from the binary vector scheme; ffl use of a various number of parents to produce an offspring; ffl development of specialized crossover operators; ffl use of local search methods to improve solutions obtained through crossover; ffl use of repair operators; ffl ....

M¨uhlenbein H., M. Gorges-Schleuter and O. Kr¨aamer (1988), "Evolution algorithms in combinatorial optimization", Parallel Computing 7, 65--88.

....in [3] were over local minima obtained by a randomized implementation of the 2 Opt local search heuristic. The current study augments [3] by using four additional local search heuristics for an instance with a known globally optimal tour. We also note that other authors such as M uhlenbein et al. [15] and Sourlas [18] have used similar plots to justify their heuristics. However, our results in [3] and in this report (i) involve more solutions and use better local search heuristics; ii) compare mean distances to other solutions, in addition to distances to the optimal solution; iii) lead to ....

H. M¨uhlenbein, M. Georges-Schleuter, and O. Kr¨amer, "Evolution Algorithms in Combinatorial Optimization, " Parallel Computing 7, 1988, pp. 65--85.

No context found.

H. Muhlenbein, M. Gorges-Schleuter, and O. Kramer, \Evolution Algorithms in Combinatorial Optimization," Parallel Computing, 7, (1988), 65-88.

No context found.

H. Muhlenbein, M. Gorges-Schleuter, and O. Kramer. Evolution algorithms in combinatorial optimization. Parallel Comput., 7:65-85, 1988.

No context found.

Muhlenbein, H., Gorges-Schleuter, M. and Kramer, O. (1988). Evolution Algorithms in Combinatorial Optimization, Parallel Computing, 7, pp. 65-85.

No context found.

H. M¨uhlenbein, M. Georges-Schleuter, and O. Kr¨amer, "Evolution Algorithms in Combinatorial Optimization," Parallel Computing 7 (1988), pp. 65--85.

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