论文信息 - Towards Robustness in Neural Network Based Fault Diagnosis

Towards Robustness in Neural Network Based Fault Diagnosis

This paper studies fault-tolerance problem of feed forward neural networks implemented in pattern recognition. Based on dynamical system theory, two concepts of pseudo-attractor and its region of attraction are introduced. A method estimating fault tolerance of feed forward neural networks has been developed. This paper also presents definitions of terminologies and detailed derivations of the methodology. Some preliminary results of case studies using the proposed method are shown. Comparing to traditional methods, the proposed method has provided a framework and an efficient way for direct evaluation of fault-tolerance in feed forward neural networks.

[1] Thomas P. Trappenberg,et al. Selecting inputs for modeling using normalized higher order statistics and independent component analysis , 2001, IEEE Trans. Neural Networks.

[2] Rainer Stotzka,et al. Neural chip SAND/1 for real time pattern recognition , 1998 .

[3] Amir F. Atiya,et al. How initial conditions affect generalization performance in large networks , 1997, IEEE Trans. Neural Networks.

[4] Chong-Ho Choi,et al. Input feature selection for classification problems , 2002, IEEE Trans. Neural Networks.

[5] Zhao Yang Dong,et al. An adaptive neural-wavelet model for short term load forecasting , 2001 .

[6] Atam P. Dhawan,et al. Use of genetic algorithms with backpropagation in training of feedforward neural networks , 1993, IEEE International Conference on Neural Networks.

[7] Dit-Yan Yeung,et al. Constructive neural networks as estimators of bayesian discriminant functions , 1993, Pattern Recognit..

[8] H. H. Yan,et al. Design of a binary neural network for security classification in power system operation , 1991, 1991., IEEE International Sympoisum on Circuits and Systems.