Knapsack Problems Revisited

• Main Authors: Yi-Lun Chien, 簡逸倫
• Other Authors: Eiji Mizutani
• Format: Others
• Language: en_US
• Published: 2011
• Online Access: http://ndltd.ncl.edu.tw/handle/12390272395432700575

碩士 === 國立臺灣科技大學 === 工業管理系 === 99 === The Knapsack Problems family is one of the most important problems in Optimization fields.It was classified into NP-hard problem.Unbounded Knapsack Problem is one variant that has special properties can be efficiently exploited.In this thesis, we first give a quick scan of Knapsack Problems family, and describe these properties in details. Further, the conventional optimization method - Dynamic Programming- is conscientiously reviewed from the basic to the advance versions of this problem.Although the problem-specified Genetic Algorithm for UKP received few attentions previously, we capture the main features among the literatures and provide a simplified scheme of GA for the faster running time. According to the series of experiments, we point out the limitation of data sets which were widely applied in the literatures (especially in papers which solving UKP by GA), and discuss the trade-off between the exact algorithm DP and the approximate approach GA. Due to the powerful properties of UKP, we find out that the GA does not take many advantages comparing with DP in the problems which are usually considered large-size in UKP.