前饋式類神經網路視窗程式設計暨在化工數據的應用

Windows Programming of Feedforward Neural Network and Application in Chemical Engineering

莊興裕(Sing-Yu Jhuang)

指導教授 ： 林顯聖

國立臺北科技大學/工程學院/化學工程所/碩士(2005年)

摘要

類神經網路目前在很多領域中都有被應用的案例，而研究人員常用它做數據的擬合。根據資料複雜度的不同，所使用的網路結構也會有所不同，其中前饋式類神經網路是一種較常被使用的類型。但目前較常見的類神經網路軟體，都是屬於商業軟體，售價並不便宜，對於初學者而言是一個很大的負擔。因此本實驗室希望開發一個前饋式類神經網路的視窗化程式，並免費提供給類神經網路學習者使用。

本研究使用Visual Basic 6.0進行視窗程式的開發。程式中網路參數有兩種訓練法則可以使用，分別為動量修正逆傳遞法則與可變學習速率逆傳遞法則，並且使用Nguyen-Widrow method計算配重值與偏移量之初始值的範圍。

本研究所開發之前饋式類神經網路視窗程式，在結構上可以選擇多層的隱藏層，且轉換函數有線性函數、S形函數、雙曲正切函數、高斯函數等十種可做選擇，比實驗室開發的前一個版本，在功能上又提昇了不少。而使用此前饋式類神經網路視窗程式進行數個模擬範例的驗證，結果顯示開發的視窗程式具有很好的實用價值。

關鍵字

前饋式類神經網路 ； 視窗程式設計 ； 數據擬合

並列摘要

Neural network has been applied in many fields. Many researchers use neural network for data-fitting work in their studies. The frequently used network is feedforward neural network. Most of neural network programs are commercial software. It is a large burden for beginners in price. Hence, our laboratory wants to develop a window-based program of feedforward neural network and provide it for the beginners.

Visual Basic 6.0 is used to develop window-based program. There are two learning rules, which are momentum modification to backpropagation algorithm and variable learning rate backpropagation algorithm in the program. Furthermore, the ranges of initial weights and biases are calculated by Nguyen-Widrow method.

In the program, user can choose multiple hidden layers and ten transfer functions for neuron including Linear transfer function, Sigmoid transfer function, Hyperbolic Tangent transfer function, Gaussian transfer function, etc. The capability of the program is better than that of program developed by our laboratory before. Besides, we make some examples for the program in this thesis. These results show that the program is practicable.

並列關鍵字

feedforward neural network ； Windows programming ； data-fitting

參考文獻

[2] McCulloch, W. and Pitts, W., "A logical calculus of the ideas immanent in nervous activity," Bulletin of Mathematical Biophysics, vol. 5, 1943, pp. 115-133.

[3] Hebb, D. O., "The Organization of Behavior," New York: Wiley, 1949.

[4] Rosenblatt, F., "The perceptron: A probabilistic model for information storage and organization in the brain," Psychological Review, vol. 65, 1958, pp. 386-408.

[7] Hopfield, J. J., "Neural networks and physical systems with emergent collective computational abilities," Proceedings of the National Academy of Sciences, vol. 79, 1982, pp. 2554-2558.

[8] Rumelhart, D. E. and McClelland, J. L., eds., "Parallel Distributed Processing: Explorations in the Microstructure of Cognition," vol. 1, Cambridge, MA: MIT Press, 1986.