The Journal of Korean Institute of Communications and Information Sciences (한국통신학회논문지)

The Korean Institute of Commucations and Information Sciences (한국통신학회)
권호 표지

A new modular neural network training algorithm for step-like discontinuous function approximation

계단형 불연속 함수의 근사화를 위한 새로운 모듈형 신경회로망 학습 알고리즘

  • Published : 1997.12.01
Download PDF
⟨ Previous Next ⟩

Abstract

Theoretically, a multi-layered feedforward network has been known to be able to approximate a continuous function to an arbitrary degree of accuracy. However, these networks fail to approximate discontinuous functions when they are trained by well-known training algorithms. This paper presents a training algorithm which doesn't work consists of one or more modules, which are trained in a sequential order within subspaces of the input space, and is trained very rapidely once all modules are trained and merged. The experimantal results of applying this method indicates the proposed training algorithm is superior to traditional ones such as baskpagation.

Keywords

References

  1. Proceedings of IEEE v.76 Ill-posed problems in early vision M. Bertero;T. Poggio;V. Torre
  2. Math. Control, Signals and Systems v.2 Approximation by superpositions of a sigmoidal function C. Cybenko
  3. From images to Surfaces:A computational study of the human early visual system W. Grimson
  4. Technical Report v.908 The coupled depth/slope aporoach to surface reconstruction J. G. Harris
  5. Neurocomputing R. Hecht-Nielsen
  6. Proceedings of the International Conference on Neural Network Theory of the back-propagation neural network R. Hecht-Nielsen
  7. Sciences v.235 Computing with neural circuits:A model J. Hopfield;D. Tank
  8. Neural Networks v.2 Multilayer feedforward networks are universal approximators K. Hornick;M. Stinchcombe;H. White
  9. Proc. Natl. Acad. Sci. v.83 Analog "neuronal" networks in early vision C. Koch;J. Marroquin;A. Yuille
  10. Doki. Akad. Nauk USSR v.114 On the representation of continuous functions of many variables by superposition functions of many variables by supperposition of continuous functions of one variable and addition A. N. Kolmogorov
  11. Neural Information Processing Systems How neural nets work A. Lapedes;R. Farber;D. Anderson(ed.)
  12. Proceedings of World Congress of Neural Networks Regularization techniques in artificial neural networks J. Mao;A. K. Jain
  13. J. of the American Statistical Association v.83 no.897 Probabilistic solution of ill-posed problems in computational vision J. Marroquin;S. Mitter;T. Poggio
  14. Advances in Neural Information Processing System 4 The effective number of parameters: An analysis of generalization and regularization in nonlinear learning systems J. Moody;J. Moody(ed.);S. Hanson(ed.);R. Lippmann(ed.)
  15. Nature v.317 Computational vision and regularization theory T. Poggio;V. Torre;C. Koch
  16. Parallel Distributed Processing:Exploration in the microstructure of cognition v.1 Learning internal representations by error propagation D. Rumelhart;G. Hinton;R. Williams
  17. Proceedings of conference on Information Science and Systems, Johns Hopkins University Capabilities of four-vs. three-layer nets, and control applications E. Sontag
  18. IEEE Trans. on Circuits and Systems v.33 no.5 Simple D. Tank;J. Hopfield
  19. Solutions of ill-posed problems A. Tikhonov;V. Aresenin