Logo image
霍普菲爾網路法求解二階線性規劃問題
Thesis

霍普菲爾網路法求解二階線性規劃問題

藍坤銘
Masters, National Tsing Hua University
2007

Abstract

霍普菲爾類神經網路二階線性規劃問題多階規劃問題供應鏈規劃 Hopfield neural networksbilevel linear programming problemsmultilevel programming problemssupply chain planning
This study exploits Hopfield neural networks (HNN) with branch and bound (B&B) tree and tabu strategy for solving bilevel linear programming and bilevel linear decentralized programming (BLP) problems, which are the special cases of multilevel linear programming (MLP) problems. MLP is a useful model to manage a decentralized planning process of hierarchical organizations in the real world. The model includes both levels having conflicting goals and separated controlled decisions which are difficult to obtain an optimal solution. The newly developed HNN approaches are efficient tools to manipulate the optimization model with parallel processing, and are especially suitable for large size problems. Therefore, two types of hybrid HNN approaches, with a B&B tree or Tabu strategy, are developed to attack BLP problems. In addition, some discussions on parameter settings, transfer functions, learning rates, and etc. are also discussed for ease of using HNN approaches. Finally, some typical examples are demonstrated and an application of supply chain planning is also investigated.

Metrics

1 Record Views

Details

Logo image