论文信息 - A Measure of Noise Immunity for Functional Networks

A Measure of Noise Immunity for Functional Networks

In this paper a study of the influence of input perturbations on a functional network is presented. A quantitative measure, related to the mean squared error degradation in presence of input noise, is introduced. This measure, based on statistical sensitivity, provides an estimation of the generalization ability and noise immunity of functional networks and lets to predict the performance degradation of a functional network. The experimental results corroborated the validity of the proposed model.

Amparo Alonso-Betanzos | Enrique F. Castillo | Oscar Fontenla-Romero | Bertha Guijarro-Berdiñas

[1] Ignacio Rojas,et al. A Quantitative Study of Fault Tolerance, Noise Immunity, and Generalization Ability of MLPs , 2000, Neural Computation.

[2] A. F. Murray,et al. Modelling weight- and input-noise in MLP learning , 1996, Proceedings of International Conference on Neural Networks (ICNN'96).

[3] Chong-Ho Choi,et al. Sensitivity analysis of multilayer perceptron with differentiable activation functions , 1992, IEEE Trans. Neural Networks.

[4] Enrique F. Castillo,et al. Functional networks with applications: a neural-based paradigm [Book Review] , 1999, IEEE Transactions on Neural Networks.

[5] Eduardo Ros,et al. A New Measurement of Noise Immunity and Generalization Ability for MLPs , 1999, Int. J. Neural Syst..

[6] L. Glass,et al. Oscillation and chaos in physiological control systems. , 1977, Science.

[7] J. I. Minnix. Fault tolerance of the backpropagation neural network trained on noisy inputs , 1992, [Proceedings 1992] IJCNN International Joint Conference on Neural Networks.

[8] Christopher M. Bishop,et al. Neural networks for pattern recognition , 1995 .

[9] Christopher M. Bishop,et al. Current address: Microsoft Research, , 2022 .