Feedforward Nets for Interpolation and Classification

@article{Sontag1992FeedforwardNF,
  title={Feedforward Nets for Interpolation and Classification},
  author={Eduardo D. Sontag},
  journal={J. Comput. Syst. Sci.},
  year={1992},
  volume={45},
  pages={20-48}
}
This paper deals with single-hidden-layer feedforward nets, studying various aspects of classification power and interpolation capability. In particular, a worst-case analysis shows that direct input to output connections in threshold nets double the recognition but not the interpolation power, while using sigmoids rather than thresholds allows doubling both. For other measures of classification, including the Vapnik-Chervonenkis dimension, the effect of direct connections or sigmoidal… CONTINUE READING
Highly Influential
This paper has highly influenced 10 other papers. REVIEW HIGHLY INFLUENTIAL CITATIONS

From This Paper

Figures, tables, and topics from this paper.

Citations

Publications citing this paper.
Showing 1-10 of 102 extracted citations

Learning with recurrent neural networks

View 5 Excerpts
Highly Influenced

Learning Pattern Classification - A Survey

IEEE Trans. Information Theory • 1998
View 9 Excerpts
Highly Influenced

References

Publications referenced by this paper.
Showing 1-10 of 13 references

On the capabilities of multilayer perceptrons

E. B. Baum
J.Complexity 4(1988): 193- 215. • 1988
View 5 Excerpts
Highly Influenced

Backpropagation separates where perceptrons do

E. D. Sontag, H. J. Sussmann
Neural Networks, 4(1991): 243-249. 25 θ θ θ INPUTS { Possible direct connections OUTPUT Hidden Layer Figure 1: Feedforward Net. x x x x x O O O ? Figure 2: Labeled Samples. 26 0 1 -20 0 20 Figure 3: Sigmoid leading to infinite VC dimension. 27 • 1991
View 1 Excerpt

Feedback Stabilization Using Two-Hidden-Layer Nets

1991 American Control Conference • 1991
View 2 Excerpts

Neural networks for exact matching of funtions on a discrete domain

Y. Shrivastava, S. Dasgupta
Proc. IEEE Conf. Dec. and Control , Honolulu, Dec. 1990, pp. 1719-1724. • 1990
View 1 Excerpt

Why two hidden layers and better than one

D. Chester
Proc. Int. Joint Conf. on Neural Networks , Washington, DC, Jan. 1990, IEEE Publications, 1990, p. I.265-268. • 1990
View 3 Excerpts

Similar Papers

Loading similar papers…