仿螞蟻群聚最佳化神經網路於非線性迴歸之初步探討

鄭又新(Yu-Hsin Cheng)

指導教授 ： 張江南

國立中央大學/工學院/光機電工程研究所/碩士(2007年)

摘要

由類神經網路模式提出一種非線性迴歸的方法，該方法能有效地解決傳統多元迴歸分析難以解決的非線性多元迴歸問題。
　　本文中，嘗試利用混合蟻群最佳化和傳統倒傳遞學習演算法的方式建立一種新的神經網路學習演算法。使用這樣的學習規則，能正確地計算出神經網路適當的權值增加量。期望其可以讓神經網路學習的速度有顯著地提升，也能有效地改進其系統的準確度。而為了展示這樣學習技術的優點，設計幾個非線性系統問題做為試驗。並以原先的倒傳遞學習方式與其作為比較。從模擬結果可知此一學習技術較傳統倒傳遞學習方式效率佳。

關鍵字

非線性迴歸 ； 類神經網路 ； 蟻群最佳化

並列摘要

A nonlinear multiple regression analysis method based on artificial neural network is developed. The method can resolve nonlinear multiple regression problems effectively, when traditional multiple regression analysis may not capable of doing the job.
　　In this thesis, a new learning algorithm of neural network is developed by using ant colony optimization and back-propagation (BP) learning algorithm. Based on this learning algorithm, the appropriate weight increments of neural network can be computed and decided. Therefore, not only the learning of neural network can be greatly speeded, also the accuracy of neural network''s performance can be effectively improved. For demonstrating the superiority of learning technique, several nonlinear system identification problems were tested. For comparison, same experiments were also performed by network with pure BP learning rule. From the experimental results, the learning technique we developed obviously has better performance as desired.

並列關鍵字

nonlinear regression ； neural network ； ant colony optimization

參考文獻

[16] 楊金聲,“利用類神經網路與線性迴歸進行成本預測之研究-以印刷電路板產業為例”,中原大學資訊管理學系碩士論文,2005.

[6] Amr Radi, and Riccardo Poli, “Genetic Programming Discovers Efficient Learning Rules for the Hidden and Output Layers of Feedforward Neural Networks.”, The University of Birmingham ,1998.

[7] Shynk, J.J. and Roy, S. “The LMS algorithm with momentum updating.”, IEEE International Symposium, on Circuits and Systems,1988.

[8] Lari-Najafi Hossein, Mohammed Nasiruddin, and Tariq Samad. “Effect of Initial Weight on Back-Propagation and Its Variations.”, IEEE Transactions on Systems, Man, and Cybernetics, Vol. 1, P218-219,1989.

[9] John K. Kruschke, and Javier R. Movellan, “Benefits of Gain: Speeded Learning and Minimal Hidden Layers in Back-Propagation Networks.”, IEEE Transactions on Systems, Man, and Cybernetics, Vol. 21, No.1, P273-280,1991.