Abstract
This paper concerns the initialization problem of the training algorithm in Neural Networks. We focus herein on backpropagation networks with one hidden layer. The initialization of the weights is crucial; if the network is incorrectly initialized, it converges to local minima. The classical random initialization therefore appears as a very poor solution. If we were to consider the Taylor development of the mapping problem and the nonlinearity of sigmoids, the improvements could be very significant. We propose a new initialization scheme based on the search for an explicit approximate solution to the problem of mapping between pattern and target. Simulation results are presented which show that these original initializations avoid local minima, reduce training time, obtain a better generalization and estimate the network's size.
Similar content being viewed by others
References
P. Costa, P. Larzabal and H. Clergeot, “On Initialization of NeuralNetwork”, Proceedings of ICNN 95, Perth, Australia, 29 Nov.-1 Dec., 1995.
D. Nguyenand B. Widrow, “Improving the Learning Speed of 2-Layer Neural Networks by choosing Initial Values of the Adaptative Weights”, Proc. Intl. Joint Conf. on Neural Networks, San Diego, Calif., June 1990.
S. Osowski, “New approach to selection of initial values of weights in neural functionapproximation”, Electronics Letters, Vol. 29, No. 3, pp. 313–315, Feb. 1993.
T. Denoeux and R. Lengellé, “Initializing back propagation networks with prototypes”, Neural Networks, Vol. 6, pp. 351–363, 1993.
G. Cybenko, “Approximation bysuperpositions of a sigmoidal function”, Mathematics of Control, Signal and Systems, Vol. 2, pp. 303–314, 1989.
Author information
Authors and Affiliations
Rights and permissions
About this article
Cite this article
Costa, P., Larzabal, P. Initialization of Supervised Training for Parametric Estimation. Neural Processing Letters 9, 53–61 (1999). https://doi.org/10.1023/A:1018671912219
Issue Date:
DOI: https://doi.org/10.1023/A:1018671912219