-
[show abstract]
[hide abstract]
ABSTRACT: Penalty functions are often used to handle constraints for constrained optimization problems in evolutionary algorithms. However it is difficult to control penalty parameters. To overcome this shortcoming, a new penalty function with easily-controlled penalty parameters is designed in this paper. The fitness function defined by this penalty function can distinguish feasible and infeasible solutions effectively. Meanwhile, the orthogonal design is used to generate initial population and design crossover operator. Based on these, a new genetic algorithm for constrained optimization problems is proposed. The simulations are made on five widely used benchmark problems, and the results indicate the proposed algorithm is effective.
Machine Learning and Cybernetics, 2005. Proceedings of 2005 International Conference on; 09/2005
-
[show abstract]
[hide abstract]
ABSTRACT: Crossover is a fundamental operator in genetic algorithms, through which not only an existing schema may be either eliminated or survived, but also a new schema is constructed via other existing schemata. Unfortunately the traditional schema theorem (Holland, J.H., 1975) does not take into account the positive effects of a schema construction through crossover operation. Recently, some works have been done by considering the schema construction, but they could not well characterize the evolution of a schema via crossover. We propose a new representation of a schema, called ternary representation, through which the survival and construction probabilities of a schema are given out. Eventually, we present a new improved schema theorem that considers both schema survival and construction in a uniform crossover.
Intelligent Sensors, Sensor Networks and Information Processing Conference, 2004. Proceedings of the 2004; 01/2005
-
[show abstract]
[hide abstract]
ABSTRACT: The traditional crossover operator performs the constant crossover between two parents without considering their homogeneity. Actually, the more homogeneous the parents are, the more disruptive the crossover should be. In this paper, a self-adaptive mechanism named adaptive recombination with three sub-populations (ARTS) is therefore presented to control the crossover operator of a genetic algorithm. The ARTS allows the crossover to be dynamically switched among two-point crossover (i.e., the least disruptive crossover), uniform crossover with probability 0.2, and uniform crossover with probability 0.5 (i.e., the most disruptive crossover). The experiments have shown the promising results.
Machine Learning and Cybernetics, 2004. Proceedings of 2004 International Conference on; 09/2004
-
[show abstract]
[hide abstract]
ABSTRACT: Traveling salesman problem (TSP for short) is a class of NP-hard combinatorial optimization problems. It is of great importance in both theory and applications. First, a new and simple encoding scheme is designed. For TSP with (n+1) cities, the encoding scheme is to use an integer in [0,n!-1] to represent a valid tour and each integer is corresponding to unique valid tour, and vice versa. Thus, TSP can be transformed into a problem in which we shall look for an integer in [0,n!-1] such that the length of the corresponding tour is shortest. The most obvious advantage is that it is very easy to design easy-executed and efficient crossover and mutation operators. Second, a novel local search scheme is integrated into the crossover and mutation operators to enhance their ability of exploration. Based on these, a novel and effective evolutionary algorithm for TSP is proposed and its convergence to global optimal solution with probability one is proved. At last, the numerical experiments are made for five standard test problems. The best solutions found by the proposed algorithm are better than or equal to those found by the compared algorithms. These results indicate the proposed algorithm is efficient.
Machine Learning and Cybernetics, 2004. Proceedings of 2004 International Conference on; 09/2004
-
[show abstract]
[hide abstract]
ABSTRACT: A new evolutionary algorithm for degree-constrained minimum spanning tree problem, which is an NP-complete problem, is proposed, and Prufer encoding method is adopted to encode the solutions of the problem. To enhance the algorithm, two new genetic operators are specifically designed, which can avoid producing infeasible solutions. In addition, two novel local search operators are designed and combined into the algorithm to further improve the quality of solutions. As a result, the proposed evolutionary algorithm can efficiently explore the search space.
Computational Intelligence and Multimedia Applications, 2003. ICCIMA 2003. Proceedings. Fifth International Conference on; 10/2003
