Wyniki wyszukiwania - BazTech

1

Multi-combinative strategy to avoid premature convergence in genetically-generated fuzzy knowledge bases

Achiche S., Balazinski M., Baron L.

Journal of Theoretical and Applied Mechanics

|

2004

|

Vol. 42 nr 3

417-444

EN

A growing number of industrial fields is concerned by complex and multiobjective problems. For this kind of problems, optimal decision making is critical. Decision support systems using fuzzy logic are often used to deal with complex and large decision making problems. However the main drawback is the need of an expert to manually construct the knowledge base. The use of genetic algorithms proved to be an effective way to solve this problem. Genetic algorithms model the life evolution strategy using the Darwin theory. A main problem in genetic algorithms is the premature convergence, and the last enhancements in order to solve this problem include new multi-combinative reproduction techniques. There are two principal ways to perform multi-combinative reproduction within a genetic algorithm, namely the Multi-parent Recombination, Multiple Crossover on Multiple Parents (MCMP); and the Multiple Crossover Per Couple (MCPC). Both techniques try to take the most of the genetic information contained in the parents. This paper explores the possibility to decrease premature convergence in a real/binary like coded genetic algorithm (RBCGA) used in automatic generation of fuzzy knowledge bases (FKBs). The RBCGA uses several crossover mechanisms applied to the same couple of parents. The crossover are also combined in different ways creating a multiple offspring from the same parent genes. The large family concept and the variation of the crossovers should introduce diversity and variation in otherwise prematurely converged populations and hence, keeping the search process active.

PL

Rosnącej liczbie dziedzin, którymi zainteresowany jest przemysł, towarzyszą złożone zagadnienia wieloobiektowe. Dla takich zagadnień optymalne podejmowanie decyzji jest krytyczne. Często dla wsparcia procesu decyzyjnego w złożonych problemach stosuje się układy logiki rozmytej. Kłopotem pozostaje jednak potrzeba manualnego wygenerowania bazy wiedzy poprzez eksperta. Okazuje się, że pewnym rozwiązaniem tego problemu może być użycie algorytmów genetycznych. Algorytmy takie modelują zagadnienie ewolucyjne na podstawie teorii Darwina. Głównym problemem w algorytmach genetycznych jest przedwczesna konwergencja, której próby wyeliminowania oparto na strategii multikombinowanych technik reprodukcji. Występują zasadniczo dwie drogi realizacji techniki reprodukcji: Multiple Crossover on Multiple Parents (MCMP) oraz Multiple Crossover Per Couple (MCPC). Obydwie metody celują w wykorzystanie jak największej ilości informacji genetycznej od rodziców. W artykule zajęto się możliwością ograniczania przedwczesnej konwergencji w rzeczywistym/binarnym kodzie genetycznym (RBCGA) używanym w automatycznymgenerowaniu rozmytych baz wiedzy (FKBs). Algorytm RBCGA stosuje kilka mechanizmów krzyżowania genów w odniesieniu do tej samej pary rodziców. Mechanizmy te przeróżnie kombinowane pozwalają na wielokrotną kreację potomstwa od tej samej pary rodziców. Koncepcja dużej rodziny i różnicowanie krzyżowania powinny wprowadzić dywersyfikację nowogenerowanych pokoleń, które w przeciwnym razie szybko uległyby konwergencji. Zapobieżenie temu zjawisku poprzez strategię multikombinacyjną utrzymuje proces poszukiwania rozwiązania w stanie aktywnym.

2

Advances in parallel heterogeneous genetic algorithms for continuous optimization

Alba E., Luna F., Nebro A. J.

International Journal of Applied Mathematics and Computer Science

|

2004

|

Vol. 14, no 3

317-333

EN

In this paper we address an extension of a very efficient genetic algorithm (GA) known as Hy3, a physical parallelization of the gradual distributed real-coded GA (GD-RCGA). This search model relies on a set of eight subpopulations residing in a cube topology having two faces for promoting exploration and exploitation. The resulting technique has been shown to yield very accurate results in continuous optimization by using crossover operators tuned to explore and exploit the solutions inside each subpopulation. We introduce here a further extension of Hy3, called Hy4, that uses 16 islands arranged in a hypercube of four dimensions. Thus, two new faces with different exploration/exploitation search capabilities are added to the search performed by Hy3. We analyze the importance of running a synchronous versus an asynchronous version of the models considered. The results indicate that the proposed Hy4 model overcomes the Hy3 performance because of its improved balance between exploration and exploitation that enhances the search. Finally, we also show that the async Hy4 model scales better than the sync one.

3

A generic evolutionary computation approach based upon genetic algorithms and evolution strategies

Affenzeller M.

Systems Science

|

2002

|

Vol. 28, nr 2

59-71

EN

Many problems that are treated by genetic algorithms (GA) belong to the class of NP-complete problems. GA are frequently faced with a problem similar to that of stagnating in a local but not global solution. This drawback, called premature convergence, occurs when the population of a GA reaches such a suboptimal state that the genetic operators can no longer produce offspring that outperforms their parents. The author considers GA as artificial self-organizing processes in a bionically inspired generic way. In doing so he introduces an advanced selection model for GA that allows adaptive selective pressure handling in a way that is quite similar to evolution strategies. This enhanced GA model allows further extensions like the introduction of a concept to handle multiple crossover operators in parallel or the introduction of a concept of segregation and reunification of smaller subpopulations. Both extensions rely on a variable selective pressure. The experimental part of the paper discusses the new algorithms for the traveling salesman problem (TSP) as a well documented instance of a multimodal combinatorial optimization problem achieving results which significantly outperform the results obtained with a contrastable GA.