政大學術集成


Please use this identifier to cite or link to this item: https://ah.nccu.edu.tw/handle/140.119/100251


Title: Dynamic Programming Variant in Evolution Strategies for Production Allocation Problems
Other Titles: 應用動態規劃式演化策略解決生產分配問題
Authors: 侯永昌;張應華
Hou, Young-Chang;Chang, Ying-Hua
Keywords: 演化策略;生產分派問題;動態規劃式演化策略;路徑編碼法;路徑突變法
Evolution Strategies;Production Allocation Problems;DP Variant Evolution Strategies;Path Encoding Method;Path Mutation
Date: 2003-06
Issue Date: 2016-08-16 15:16:06 (UTC+8)
Abstract: 演化策略(evolution strategies)通常應用於目標變數為實數值的最佳化問題上,其主要是利用突變(mutation)的方式於解答空間中搜尋出最好的滿足解,利用類似的演化過程有許多組合最佳化的問題合適且成功地被解決,如二次指派問題(quadratic assignment problems),工程設計最佳化問題(engineering design optimization problems)等等,但一般的演化計算技術應用於這些問題時,將會浪費很多的時間於處理在演化過程中獲得無效解的情形上,以致於造成演算法效率不彰。本研究提出一種新的染色體編碼方式,稱為路徑編碼法(the path-encoding method),其修改動態規劃的路徑搜尋概念於組合演化策略(combinatorial evolution strategies)上,以提高其演化效能,並以NP-hard的生產分派問題(production allocation problems)來當成測試的實例,在實驗中對路徑編碼法、組合編碼法(combination-encoding method)、懲罰編碼法(penalty-encoding method)和整數規劃法(integer programming)等四種方法做一比較,結果顯示出本研究所提之路徑編碼法效果最好,其次是組合編碼法,再者是懲罰編碼法,最後是整數規劃法,其主要的原因是路徑編碼法可以有效地縮小搜尋解答空間的範圍,以加快收斂的速度。
Evolution strategies are applied to optimize real-valued vectors of objective variables. These strategies rely primarily on mutation to explore the solution search space. Many combinatorial optimization problems such as quadratic assignment problems, engineering design optimization problems, and others can be successfully solved by analog evolution. This evolutionary algorithm wastes much time in managing invalid solutions and is typically less efficient.This paper presents a new approach, called the path-encoding method, which modifies the path searching idea of dynamic programming for combinatorial evolution strategies to enhance the performance of evolutionary process. The NP-hard production allocation problem is used to evaluate the effectiveness of the approach. This experiment compares the proposed approach to the combination-encoding method, the penalty-encoding method and integer programming. The computed results show that the proposed approach is always feasible and outperforms the others because it narrows the solution search space.
Relation: 資管評論, 12, 1-16
MIS review
Data Type: article
Appears in Collections:[MIS review] Journal Articles

Files in This Item:

File Description SizeFormat
12-1-16.pdf1032KbAdobe PDF479View/Open


All items in 學術集成 are protected by copyright, with all rights reserved.


社群 sharing