一種動態(tài)篩選樣本的前向神經(jīng)網(wǎng)絡快速學習算法
A Fast Learning Algorithm of Feedforward Neural Networks Based on Screening Samples Dynamically
-
摘要: 從理論上討論了一類隱含層激勵函數(shù)滿足Mercer條件的前向神經(jīng)網(wǎng)絡學習問題，分析了提高網(wǎng)絡學習速度的途徑，提出了一種動態(tài)篩選樣本的前向神經(jīng)網(wǎng)絡快速學習算法。它大大提高了網(wǎng)絡學習速度，克服了傳統(tǒng)的基于梯度下降的網(wǎng)絡學習方法存在的諸多弊端。算法還具有動態(tài)確定隱含層神經(jīng)元數(shù)的自構(gòu)性優(yōu)點。文中通過具體數(shù)值試驗驗證了上述算法的可行性和優(yōu)越性。Abstract: The learning issue of feedforward neural networks whose activation function of hidden neurons satisfies Mercer condition is discussed in theory. The approach to improving learning speed is investigated. Then a fast learning algorithm of feedforward neural networks based on screening samples dynamically is proposed, which improves learning speed, solves the abuses of those learning algorithm based on gradient decent method and has the self-configuring advantage by determining the number of hidden neuron dynamically. The reliability and advantage of the proposed algorithm are illustrated concretely through test.
-
Vitaly Schetinin. A learning algorithm for evolving cascade neural networks[J].Neural Processing Letters.2003, 17(1):21-31[2]Chng E S, Chen S, Mulgrew B. Gradient radial basis function networks for nonlinear and nonstationary time series prediction[J].IEEE Trans. on Neural Networks.1996, 7(1):190-194[3]Engozinger S, Tomsen E. An accelerated learning algorithm for multiplayer perceptions: Optimization layer by layer[J].IEEE Trans. on Neural Networks.1995, 6(1):31-42[4]Scalero S, Tepedelenlioglu N. A fast new algorithm for training feedforward networks[J].IEEE Trans. on Signal Processing.1992, 40(1):202-210[5]葉軍, 張新華. 多層前向神經(jīng)網(wǎng)絡的快速學習算法及其應用. 控制與決策, 2002, 17(suppl.): 817-819.[6]劉鐵男, 段玉波, 陳廣義等. 多層前向神經(jīng)網(wǎng)絡的新型二階學習算法. 控制理論與應用, 2000, 17(5): 721-724.[7]陳亞軍. 一種多層前饋神經(jīng)網(wǎng)絡的快速學習算法. 河北師范大學學報(自然科學版), 2002, 26(6): 582-587.[8]Vapnik V. The Nature of Statistical Learning Theory. New York: Springer Verlag, 1995: 1-20.[9]張鈴. 基于核函數(shù)的SVM機與三層前向神經(jīng)網(wǎng)絡的關系. 計算機學報, 2002, 25(7): 696-700. -
計量
- 文章訪問數(shù): 2529
- HTML全文瀏覽量: 65
- PDF下載量: 634
- 被引次數(shù): 0
下載:
