Utilizing Artificial Neural Network Techniques for Solving Fuzzy Multi-level Programming Problems

• Main Authors: Han-Chyi Hsiao, 蕭翰琦
• Other Authors: Hsu-Shih Shih
• Format: Others
• Language: zh-TW
• Published: 2002
• Online Access: http://ndltd.ncl.edu.tw/handle/63044218524497443405

碩士 === 義守大學 === 工業管理學系 === 90 === This study aims to utilize the dynamic system of artificial neural networks (ANNs) to solve fuzzy multi-level programming problems (MLPPs). In this analysis, basic concepts of ANNs are discussed and an optimization problem is converted into an adequate energy function through Lagrangian multiplier and penalty function. Then, the proposed ANNs procedure is proven to be feasible for optimization. The procedure is extended to solve the multi-objective linear programming problem (MOLP) with crisp and fuzzy coefficients. After that, the procedure is adopted to deal with MLPPs with a top-down process. The issues relevant to both coefficients are also discussed. Furthermore, the procedure is verified through a network design example. The algorithm of ANNs has been viewed as a computational technique since Hopefield and Tanks’ work (1985). It enables the transfer of the optimization problem into a system of non-linear differential equations based on an energy function. When the dynamic system reaches a steady state, the optimal solution can be obtained. The non-tradition algorithm is efficient for solving complex problems, and is especially useful for implementation on a very-large-scale-integrated (VLSI), in which the MLPPs can be solved on a real time basis.