用于神經(jīng)網(wǎng)絡容錯的動態(tài)冗余BP算法
A DYNAMIC REDUNDANCY BP ALGORITHM APPLIED IN THE FAULT TOLERANCE OF NEURAL NETWORKS
-
摘要: 多層感知器(MLP)的容錯性傳統(tǒng)上采用改進算法和部件冗余方法。該文提出了一種動態(tài)冗余BP算法,這種方法在傳統(tǒng)的帶沖量項的自適應BP算法的學習過程中,根據(jù)各權(quán)值重要度的不同選取重要的權(quán)值進行冗余處理。該算法能有效地提高網(wǎng)絡的容錯能力,與學習中注入故障這一典型的容錯改進算法相比,盡管容錯能力并不突出,但相對可節(jié)省大量的學習時間。
-
關(guān)鍵詞:
- 多層前向神經(jīng)網(wǎng)絡; 容錯; 冗余; 動態(tài)冗余
Abstract: There are two main types of approaches in the research of fault tolerance of Multilayer Perceptrons(MLP): improvement in the learning algorithm and component redundancy after training. A dynamic redundancy BP algorithm is presented. In the training steps of the conventional adaptive BP algorithm with a momentum term, the most important weights are replicated based on their significance. Applying the algorithm the fault tolerance of a network can be improved effectively. Compared with fault injection while training--a typical improved learning algorithm, although this dynamic redundancy algorithm gives no prominence in fault tolerance, the training time can be greatly reduced. -
A.F. Murray, P. J. Edwards, Enhanced MLP performance and fault tolerance resulting from synaptic weight noise during training, IEEE Trans. on Neural Networks, 1994, NN-5(5), 792802.[2]P. Kerlinzin, P. Refregier, Theoretical investigation of the robustness of multilayer perceptrons,Analysis of the linear case and extension to nonlinear networks, IEEE Trans. on Neural Networks.1995, NN-6(3), 560-571.[3]B.S. Arad, A. El-Anway, On fault tolerant training of feedforward neural networks, Neural Networks, 1997, 10(3), 539-557.[4]D.S. Phatak, I. Koren, Complete and partial fault tolerance of feedforward neural nets, IEEE Trans. on Neural Networks, 1995, NN-6(2), 446-456.[5]D.S. Phatak, I. Koren, Fault tolerance of feedforward neural nets for classification tasks, IJCNN,Nagoya, Japan, 1992, II-386-391.[6]Xu Liqin.[J].Hu Dongcheng, A new component redundancy method in MLPs fault tolerance, ICEMI99, Harbin, China.1999,:- -
計量
- 文章訪問數(shù): 2041
- HTML全文瀏覽量: 118
- PDF下載量: 426
- 被引次數(shù): 0
下載:
