Summary: | 碩士 === 逢甲大學 === 資訊工程學系 === 89 === Matrix can be used to describe many kinds of problems, for example the distribution, assignment, and facility location problems. Directly adopting matrix as the representation for chromosome encoding, the evolutionary algorithm is shown to be the least complicated and efficient way for solving this type of problems. In this article, the author proposes a newly developed matrix-based mutate operator which provides a good solution for problems dealing with large matrix size. The method successfully combines a high efficiency screening mechanism of orthogonal array and divide-and-conquer strategy to handle large dimension matrix. As a norm of the matrix problems, the transportation problem is used to test the efficiency of the new algorithm by comparing with other existing algorithms. It is demonstrated that the proposed intelligent evolutionary programming is superior for large dimension problems, both for the linear and non-linear transportation problems, though less improvement is achieved when used to solve relative small matrix problem. Furthermore, the proposed strategy has the potential in broad applications for other matrix problems since heuristic is not used in the algorithm.
|