Skip to main content
SpringerLink
Log in

A neuron model with trainable activation function (TAF) and its MFNN supervised learning

  • Scientific Papers
  • Published:
Science in China Series : Information Sciences Aims and scope Submit manuscript

Abstract

This paper addresses a new kind of neuron model, which has trainable activation function (TAF) in addition to only trainable weights in the conventional M-P model. The final neuron activation function can be derived from a primitive neuron activation function by training. The BP like learning algorithm has been presented for MFNN constructed by neurons of TAF model. Several simulation examples are given to show the network capacity and performance advantages of the new MFNN in comparison with that of conventional sigmoid MFNN.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Log in via an institution

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Institutional subscriptions

Similar content being viewed by others

References

  1. Segee, B. E., Using spectral techniques for improved performance in ANN, Proc. of IEEE Intl. Conf. on Neural Networks, 1993, 500–505.

  2. Lee, S., Kil, R. M., A Gaussian potential function network with hierarchically self-organizing learning, Neural Networks, 1991, 4: 207–224.

    Article  Google Scholar 

  3. Stork, D. G., How to solve the N-bit parity problem with two hidden units, NN, 1992, 5: 923–926.

    Google Scholar 

  4. Stork, D. G., N-bit parity problems: A reply to Brown and Korn, Neural Networks, 1993, 6: 607–609.

    Article  Google Scholar 

  5. Wu, Y. S., A new approach to design a simplest ANN for performing certain specific problems, Proc. ICONI’ 95 Beijing, 1995, 1: 477–480.

    Google Scholar 

  6. Wu, Y. S., How to choose appropriate transfer function in designing a simplest ANN for solving specific problems, Science in China, Ser. E., 1996, 39(4): 369–372.

    MATH  Google Scholar 

  7. Wu, Y. S., Zhao, M. S., Ding, X. Q., A new kind of multilayer perceptron with tunable activation function and its application, Science in China, Ser. E., 1997, 40(1): 105–112.

    Article  MATH  Google Scholar 

  8. Wu, Y. S., Zhao, M. S., Ding, X. Q., A new perceptron model with tunable activation function, Chinese Journal of Electronics, 1996, 5 (20): 55–62.

    Google Scholar 

  9. Hornik, K., Approximation capabilities of multilayer feedforward networks, NN, 1991, 4: 251–257.

    Google Scholar 

  10. Baum, E. B., Lang, K. L., Constructing hidden units using examples and queries, in Advances in Neural Information Processing System 3, San Mateo, CA (eds. Lippman, P., Moody, J. E., Touretzby, D. S. et al.), Morgan Kaufmann, 1991, 904–910.

  11. Chen, Q. C., Generating-shrinking algorithm for learning arbitrary classification, Neural Networks, 1994, 7: 1477–1489.

    Article  Google Scholar 

  12. Wu, Y. S., et al., Knowledge extraction based multilayer feedforward neural networks, Proc. ICONIP’ 98, 1630–1633.

Download references

Author information

Authors and Affiliations

  1. Department of Electronic Engineering, Tsinghua University, 100084, Beijing, China

    Wu Youshou & Zhao Mingsheng

Authors
  1. Wu Youshou
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Zhao Mingsheng
    View author publications

    You can also search for this author in PubMed Google Scholar

Corresponding author

Correspondence to Wu Youshou.

Rights and permissions

Reprints and permissions

About this article

Cite this article

Wu, Y., Zhao, M. A neuron model with trainable activation function (TAF) and its MFNN supervised learning. Sci China Ser F 44, 366–375 (2001). https://doi.org/10.1007/BF02714739

Download citation

  • Received:

  • Issue Date:

  • DOI: https://doi.org/10.1007/BF02714739

Keywords

Search

Navigation