P. J. B. Hancock. Genetic algorithms and permutation problems: a comparison of recombination operators for neural net structure specification. In D Whitley, editor, Proceedings of COGANN workshop, IJCNN, Baltimore. IEEE, 1992. 13  Home/Search   Document Details and Download   Summary   Related Articles   Check

....is not new [3] 10] 13] 26] few have explained why it is important in terms of accurate fitness evaluation. The simultaneous evolution of both architectures and weights can be summarized by Fig. 2. The evolution of ANN architectures in general suffers from the permutation problem [27] [28] or called competing conventions problem [17] It is caused by the many to one mapping from genotypes to phenotypes since two ANN s which order their hidden nodes differently may have different Fig. 1. A typical cycle of the evolution of architectures. Fig. 2. A typical cycle of the evolution of ....

P. J. B. Hancock, "Genetic algorithms and permutation problems: A comparison of recombination operators for neural net structure specification," in Proc. Int. Wkshp. Combinations of Genetic Algorithms and Neural Networks (COGANN-92), D. Whitley and J. D. Schaffer, Eds. Los Alamitos, CA: IEEE Computer Soc. Press, 1992, pp. 108--122.

....accuracy by discrete values. On the other hand, if too many bits are used, chromosomes representing large ANN s will become extremely long and the evolution in turn will become very inefficient. One of the problems faced by evolutionary training of ANN s is the permutation problem [32] [113], also known as the competing convention problem. It is caused by the Fig. 4. a) An ANN which is equivalent to that given in Fig. 3(a) b) Its binary representation under the same representation scheme. many to one mapping from the representation (genotype) to the actual ANN (phenotype) ....

....low. Some researchers thus avoided crossover and adopted only mutations in the evolution of architectures [45] 128] 149] 179] 185] 197] 217] 223] although it has been shown that crossover may be useful and important in increasing the efficiency of evolution for some problems [48] [113], 212] 229] Hancock [113] suggested that the permutation problem might not be as severe as had been supposed with the population size and the selection mechanism he used because The increased number of ways of solving the problem outweigh the difficulties of bringing building blocks ....

[Article contains additional citation context not shown here]

P. J. B. Hancock, "Genetic algorithms and permutation problems: A comparison of recombination operators for neural net structure specification," in Proc. Int. Workshop Combinations of Genetic Algorithms and Neural Networks (COGANN-92),D. Computer Soc., 1992, pp. 108--122.

....is zero. It has been shown [Mayer, 1998a] that the inclusion of noncoding segments decrease the disruptive effects of crossover which might have a positive effect on ANN evolution w.r.t. the often cited problem of competing conventions (or permutation problem) However, it should be noted that Hancock (1992) explicitely reported that . simple crossover also fared well, suggesting that the permutation problem is not serious in practice [Hancock, 1992] Moreover, to our knowledge nobody has yet succeeded to give a quantitative analysis of the actual percentage of genotypes coding for identical ....

....effect on ANN evolution w.r.t. the often cited problem of competing conventions (or permutation problem) However, it should be noted that Hancock (1992) explicitely reported that . simple crossover also fared well, suggesting that the permutation problem is not serious in practice [Hancock, 1992]. Moreover, to our knowledge nobody has yet succeeded to give a quantitative analysis of the actual percentage of genotypes coding for identical phenotypes for a specific encoding. The maximum number of hidden neurons (neuron markers) has to be set in advance with this encoding scheme, hence, it ....

Hancock, P. J. B. (1992). Genetic Algorithms and permutation problems: a comparison of recombination operators for neural net structure specification. In Whitley, D., editor, Proceedings of COGANN Workshop at the International Joint Conference on Neural Networks. IEEE Press.

....parameters are optimized: time window length, context layer length and the number of hidden nodes. The mutation operator applied on these parameters randomly increases or decreases the current values in one unity. We decided not to implement crossover operators because the permutation problem [13]. This problem states that two ANNs can be functionally equivalent however bearing considerably different genotypes, which makes it difficult the generation of children with equivalent fitness. Further discussion about the permutation problem can be seen in [14] In the TR module, the networks ....

P. J. B. Hancock, Genetic Algorithms and Permutation Problems: a Comparison of Recombination Operators for Neural Net Structure Specification, In Proceedings of COGANN workshop, IJCNN, Baltimore, IEEE, 1992.

....the No Free Lunch theorem [21] says there are tasks for which the benefits of crossover outweigh its disruptiveness to neural networks. Whether your task is one of them is, alas, uncertain) Some researchers even try to invent crossover operators that cause less disruption to neural networks [9] [20] Of course, whether crossover helps or hinders depends on the other choices of operators, representation, etc. 7] as well as the attributes of the problem. This paper uses uniform crossover with some elements from Thierens [20] to reduce the disruption that crossover causes to neural ....

Peter J. B. Hancock. Genetic algorithms and permutation problems: a comparison of recombination operators for neural net structure specification. In Proceedings of the International Workshop on Combinations of Genetic Algorithms and Neural Networks, pages 108-- 122. IEEE Computer Society Press, 1992.

....same (or similar) networks may be encoded in quite different ways. The crossover of two different encodings of similar networks can result in a completely divergent NN. This broadly acknowledged problem is also known as structural functional mapping problem [27] or competing conventions problem [6]. There have been a few proposals to solve it [25, 11, 6] However, these studies also indicate that the problem has less impact than expected. GA parameters There are numerous parameters for the genetic algorithms, the neural networks and the GANN systems that have to be set: population size, ....

....different ways. The crossover of two different encodings of similar networks can result in a completely divergent NN. This broadly acknowledged problem is also known as structural functional mapping problem [27] or competing conventions problem [6] There have been a few proposals to solve it [25, 11, 6]. However, these studies also indicate that the problem has less impact than expected. GA parameters There are numerous parameters for the genetic algorithms, the neural networks and the GANN systems that have to be set: population size, mutation and crossover rates, the use of distributed GA, ....

Peter J.B. Hancock. Genetic algorithms and permutation problem: a comparison of recombination operators for neural net structure specification. In International Workshop on Combinations of Genetic Algorithms and Neural Networks, pages 108--122, Baltimore, 1992. IEEE.

....is that, as the network s size increases, the length of the chromosome increases very quickly. Where N is the number of nodes in the network, the chromosome will be N 2 bits long (or about N 2 2 in a feed forward network) Now, some experiments where direct encoding was used will be examined. Hancock (1992), who used direct encoding (or strong coding, as he calls it) notes that this method is appropriate since his concern is the precise definition of relatively small nets. Huber et al. 1995. were also successful in using this direct encoding method to construct small feed forward neural ....

....layers of neural networks. Therefore, there are multiple ways to code for the same network. In other words, there is a many to one mapping from genotype to phenotype (Liu 1996) This can make crossover of connections between networks a rather inefficient way to search a network s structure space. Hancock (1992) attempted to solve this problem by devising special crossover operators. His conclusions were that the permutations problem may not be as important a problem as had been suspected. He found that the GA was capable of finding solutions within all of the possible permutations of a network. When ....

Hancock, P.J.B. (1992). Genetic Algorithms and permutation problem: a comparison of recombination operators for neural net structure specification, In Proceedings of COGANN-92, D. Whitley and J.D. Schaffer (Eds.), IEEE, Computer Society Press. U.S.A., June 6, 1992. p. 108-122.

....Guha, Aloke, 727, 728, 729, 730] Guo, Zhichao, 719, 720] Hakkarainen, Juha, 343, 361] Hall, T. J. 637] Hamad, Denis, 586] Hamalainen, Ari, 178, 253, 10] Hamersma, H. 425] Hammer, Jurgen, 588] Han, Seung Soo, 344, 393, 459, 469, 546] Han, S. S. 78] Hancock, Peter J. B. [721, 722, 723, 724, 725, 796] Handroos, H. 543] Hanebeck, Uwe D. 345] Hansen, J. V. 346] Hansen, Kim Kortermand, 40] Hansen, L. K. 274] Hanson, Thomas, 947, 952] Happel, B. L. M. 726] Happel, Bart L. M. 41] Harget, A. 456] Haring, S. 503] Harp, Steven Alex, 727, 728, 729, 730, 731, 732, 733, ....

.... deterministic, 466] dynamic, 295] neural Darwinism, 556] neural netoworks, 243] neural network, 539] control, 280, 515] design, 444] fuzzy, 423, 477] rule extraction, 464] signal processing, 287] structure selection, 172] training, 354] wavelet, 281] neural networks, [619, 665, 666, 914, 939, 599, 600, 650, 651, 727, 783, 799, 945, 946, 601, 606, 613, 642, 649, 656, 657, 711, 728, 729, 854, 909, 947, 948, 949, 950, 951, 952, 970, 602, 607, 610, 615, 634, 637, 645, 652, 663, 670, 678, 679, 680, 681, 682, 683, 684, 685, 714, 721, 730, 756, 757, 762, 764, 770, 781, 782, 785, 786, 787, 800, 810, 815, 834, 869, 874, 875, 879, 885, 902, 904, 907, 910, 920, 923, 940, 953, 954, 971, 973, 974, 975, 976, 977, 608, 618, 620, 622, 628, 630, 644, 654, 660, 661, 664, 674, 686, 687, 688, 693, 712, 713, 731, 735, 758, 763, 793, 796, 806, 813, 817, 841, 857, 859, 860, 861, 864, 886, 889, 890, 911, 913, 915, 921, 936, 942, 955, 956, 960, 972, 978, 979, 980, 981, 982, 592, 594, 595, 605, 611, 614, 616, 621, 635, 636, 638, 639, 643, 658, 669, 671, 672, 673, 689, 690, 691, 696, 700, 716, 717, 719, 720, 722, 723, 724, 725, 726, 733, 734, 736, 737, 738, 739, 740, 741, 766, 767, 773, 774, 778, 784, 790, 791, 795, 797, 801, 804, 809, 814, 818, 819, 821, 822, 823, 824, 829, 835, 836, 837, 838, 845, 846, 858, 871, 876, 880, 881, 900, 906, 908, 917, 919, 922, 925, 926, 941, 957, 958, 965, 967, 968, 983, 984, 593, 596, 597, 598, 609, 612, 617, 629, 631, 632, 633, 640, 641, 646, 653, 659, 662, 668, 675, 677, 692, 694, 698, 701, 703, 704, 705, 706, 709, 742, 743, 744, 745, 746, 747, 748, 749, 750, 751, 752, 753, 761, 765, 768, 769, 771, 775, 776, 780, 788, 794, 802, 803, 805, 807, 808, 811, 812, 825, 827, 832, 833, 842, 844, 847, 848, 849, 850, 856, 862, 863, 865, 866, 868, 870, 872, 873, 877, 884, 887, 892, 893, 894, 895, 896, 901, 905, 918, 927, 929, 930, 931, 933, 934, 935, 937, 938, 943, 944, 961, 963, 966, 985, 986, 987, 988, 13, 14, 18, 26, 38, 40, 43, 63, 75, 84, 86, 88, 91, 96, 99, 103, 106, 108, 111, 112, 113, 114, 118, 121, 124, 125, 128, 129, 134, 137, 138, 139, 140, 146, 147, 151, 153, 156, 157, 166, 167, 169, 180, 190, 191, 192, 193, 196, 199, 200, 209, 215, 217, 222, 226, 227, 230, 236, 238, 241, 242, 244, 254, 258, 260, 264, 266, 268, 269, 271, 272, 277, 279, 283, 292, 295, 296, 297, 303, 305, 318, 321, 322, 323, 327, 333, 338, 349, 353, 366, 372, 374, 375, 378, 381, 388, 390, 398, 400, 403, 404, 407, 408, 411, 415, 427, 440, 441, 448, 453, 454, 456, 458, 459, 463, 473, 360, 475, 478, 481, 487, 489, 490, 491, 492, 493, 495, 499, 504, 505, 508, 512, 523, 525, 526, 530, 531, 533, 542, 556, 562, 566, 583] neural networks age, 206] analysis, 362] architecture, 41, 184] associative memory, 570] back propagation, 397] back propagation, 502] backpropagation, 64, 69, 115, 351, 447] Baldwin effect, 155] Bayesian, 174, 339, 527] binary logic, 565] biological, 779] Boltzmann, 70] ....

[Article contains additional citation context not shown here]

Peter J. B. Hancock. Genetic algorithms and permutation problems: A comparison of recombination operators for neural net structure specification. In Schaffer and Whitley [638], pages 108--122. y(Fogel/bib) ga:Hancock92c.

....278] Gupta, Akhil Datta, 454] Gurdal, Zafer, 285] Gutierrez, D. 360] Hadj Alouane, A. B. 82] Haftka, Raphael T. 285, 286] Hahn, S. 84] Hahn, Song Yop, 610] Hajela, Prabhat, 287] Hall, Lester C. 98] Hampo, Richard J. 288, 289] Han, Myung Mook, 290] Hancock, Peter J. B. [291, 292, 293, 294] Happel, B. L. M. 295] Harala, Sauli, 23] Harp, Steven Alex, 296, 297] Harper, T. R. 376] Hartley, Stephen J. 118] Hartmann, Uwe, 673] Harvey, Inman, 298, 299, 300, 301, 302, 303, 304, 305, 306, 307] Hatcher, W. J. 376] Hatjimihail, Aristides T. 308] Hattori, T. 664] ....

.... [62] task allocation, 701] parallel programming, 232] parallelism, 690, 693] parameter estimation, 429, 664] parameters, 88] particle physics, 95] patent, 210, 405, 406, 409, 410] path planning, 231] pattern recognition, 316, 317, 331, 346, 680] PCB design, 676] permutations, [293] pH, 377] physical chemistry, 361, 362, 363] physics, 312, 313] optics, 450] placement, 369] planning, 225, 227, 266, 655] assembly , 435] polymer folding, 362] 2D, 361] popular, 436] population size, 21, 34, 35, 248, 263] 100, 213, 317, 453] 100; 200, 266] 30, 530, ....

[Article contains additional citation context not shown here]

Peter J. B. Hancock. Genetic algorithms and permutation problems: A comparison of recombination operators for neural net structure specification. In Schaffer and Whitley [115], pages 108--122. y(Fogel/bib) ga:Hancock92c.

....tasks with many input and output nodes are rather rare. Structural Functional Mapping Problem One problem of encoding neural network information into a genome is what Whitley calls the structural functional mapping problem [Whitley, 1990, 1992] which is also refered to as permutation problem [Hancock, 1992] or competing conventions problem [Hancock, 1992] Whitley, 1992] In most encoding strategies, it is possible that the same, or nearly the same, phenotype may have quite different genotypes. The reason for this effect is rooted in the effect that the function of certain neurons is not determined ....

....rare. Structural Functional Mapping Problem One problem of encoding neural network information into a genome is what Whitley calls the structural functional mapping problem [Whitley, 1990, 1992] which is also refered to as permutation problem [Hancock, 1992] or competing conventions problem [Hancock, 1992], Whitley, 1992] In most encoding strategies, it is possible that the same, or nearly the same, phenotype may have quite different genotypes. The reason for this effect is rooted in the effect that the function of certain neurons is not determined by their position in the network, but their ....

[Article contains additional citation context not shown here]

: Peter J.B. Hancock: "Genetic Algorithms and permutation problem: a com- - 65 - parison of recombination operators for neural net structure specification", in: Proceedings of the International Workshop on Combinations of genetic algorithms and neural networks, p. 108-122, Baltimore, IEEE.

....for (near) optimal 2 . Most such methods encode the entire network description in the chromosome of the GA including the weights. Most problems with using GAs for neural network design have been because of the non robust behavior of weights during GA operations such as crossover (Hancock[Han92] for example) Since their conception, neural networks have been used not only for their phenomenal performance in the retrieval of information, but also for the novelty (at least at the time) of learning that same information. Because of the usefulness of having a paradigm that learns on its own, ....

Peter J. B. Hancock. "Genetic Algorithms and Permutation Problems: A Comparison of Recombination Operators for Neural Net Structure Specification". IEEE . (May 1992) pp. 108--122.

No context found.

P. J. B. Hancock. Genetic algorithms and permutation problems: a comparison of recombination operators for neural net structure specification. In D Whitley, editor, Proceedings of COGANN workshop, IJCNN, Baltimore. IEEE, 1992. 13

No context found.

P. Hancock. Genetic Algorithms and Permutation Problems: a Comparison of Recombination Operators for Neural Structure Specification. Proceedings of COGANN Workshop, IJCNN, 1992. IEEE Computer Society Press.

No context found.

P. J. B. Hancock, "Genetic algorithms and permutation problems: a comparison of recombination operators for neural net structure specification, " in Proc. Int. Workshop Combinations Genetic Algorithms Neural Networks (COGANN-92), D. Whitley and J. D. Schaffer, Eds., Los Alamitos, CA: IEEE Comput. Soc. Press, 1992, pp. 108--122.

No context found.

P. J. Hancock, "Genetic algorithms and permutation problems: a comparison of recombinationoperators for neural net structure specification," in Combinations of Genetic Algorithms and Neural Networks (L. D. Whitley and J. D. Schaffer, eds.), pp. 108-- 122, Los Alamitos, CA: IEEE Computer Society Press, June 1992.

No context found.

: Peter J.B. Hancock: "Genetic Algorithms and permutation problem: a comparison of recombination operators for neural net structure specification", in: Proceedings of the International Workshop on Combina- - 91 - tions of genetic algorithms and neural networks, p. 108-122, Baltimore, IEEE.

No context found.

P. J. B. Hancock. Genetic Algorithms and Permutation Problems: a Comparison of Recombination Operators for Neural Net Structure Specification. In D. Whitley, editor, Proceedings of COGANN workshop, IJCNN, Baltimore. IEEE, 1992.

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