Input Noise Immunity of Multilayer Perceptrons

  • Published : 1994.04.30

Abstract

In this paper, the robustness of the artificial neural networks to noise is demonstrated with a multilayer perceptron, and the reason of robustness is due to the statistical orthogonality among hidden nodes and its hierarchical information extraction capability. Also, the misclassification probability of a well-trained multilayer perceptron is derived without any linear approximations when the inputs are contaminated with random noises. The misclassification probability for a noisy pattern is shown to be a function of the input pattern, noise variances, the weight matrices, and the nonlinear transformations. The result is verified with a handwritten digit recognition problem, which shows better result than that using linear approximations.

Keywords

References

  1. IEEE ASSP Magazine An introduction to computing with neural nets Lippmann, R.P.
  2. IEEE Trans. Neural Networks v.1 Sensitivity of feedforward neural networks to weight errors Stevenson, M.;Winter, R.;Windrow, B.
  3. IEEE Trans. Neural Networks v.3 Analysis of the effects of quantization in multilayer neural networks using a statistical model Xie, Y.;Jabri, M.A.
  4. IEEE Trans. Neural Networks v.3 Sensitivity analysis of multilayer perceptron with differentiable activation functions Choi, J.Y.;Choi, C.H.
  5. Proc. IJCNN'92 Beijing v.II An analysis on the classification performance of multilayer perceptrons Oh, S.H.;Lee, Y.
  6. Proc. IJCNN'90 San Diego v.I Analyses of the hidden units of backpropagation model by singular value decomposition Xue, Q.;Hu, Y.;Tompkins, W.T.
  7. IEEE Trans. Neural Networks Effect of nonlinear functions on correlation between weighted sums in multilayer perceptrons Oh, S.H.;Lee, Y.
  8. IEEE Communication Magazine Information theory, complexity, and neural networks Abu-Mostafa, Y.S.
  9. Proc. IJCNN'93 Nagoya v.III Analysis on the efficiency of pattern recognition layers using information measures Lee, Y.;Song, H.K.
  10. Probability, Random variables, and Stochastic Processes Papoulis, A.
  11. Neural Networks v.5 Approximation of continuous functions on $r^d$ by linear combinations of shifted rotations of a sigmoid function with and without scaling Ito, Y.