Cascade Correlation convergence theorem
In this paper we prove a theorem which guarantees that the Cascade Correlation algorithm has at least a speed of convergence of order O(1/n h ), where n h is the number of hidden neurons, when approximating a function which is series of sigmoids with a finite number of terms.
KeywordsHide Layer Finite Number Hide Neuron Output Neuron Hide Unit
Unable to display preview. Download preview PDF.
- Barron A.R., IEEE Trans. I. T., vol.39, no.3, May 1993, pp.930–945.Google Scholar
- Drago G.P. and Ridella S., Proc. of the ICANN’93, pp.750–754.Google Scholar
- Drago G.P. and Ridella S., submittedGoogle Scholar
- Fahlman S.E. and Lebiere C., CMU-CS-90–100, February 14, 1990.Google Scholar
- Jones L.K., Annals of Statistics, vol.20, no. l, 1992, pp.608–613.Google Scholar