Abstract
Backpropagation, (BP), is one of the most frequently used practical methods for supervised training of artificial neural networks. During the learning process, BP may get stuck in local minima, producing suboptimal solution, and thus limiting the effectiveness of the training. This work is dedicated to the problem of avoiding local minima and introduces a new technique for learning, which substitutes gradient descent algorithm in the BP with an optimization method for a global search in a multi-dimensional parameter (weight) space. For this purpose, a low-discrepancy LP τ sequence is used. The proposed method is discussed and tested with common benchmark problems at the end.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Preview
Unable to display preview. Download preview PDF.
References
Baldi, P., Hornik, K.: Learning in Linear Neural Networks: a Survey, IEEE Trans. Neural Networks 6 (1995) 837–858
Bianchini, M., Gori, M.: Optimal Learning in Artificial Neural Networks: A Review of Theoretical Results. Neurocomputing 13 (1996) 313–346
Blum, E.K.: Approximation of Boolean Functions by Sigmoidal Networks: Part I: XOR and Other Two-variable Functions. Neural Computation 1 (1989) 532–540
Bratley, P., Fox, B., Niederreiter, H.: Implementation and Tests of Low-discrepancy Sequences. ACM Trans. on Modeling and Computer Simulation 2 (1992) 195–213
Cetin, B., Barhen, J., Burdick, J.: Terminal Repeller Unconstrained Subenergy Tunneling (TRUST) for Fast Global Optimization. J. Opt. Theory and Appl. 77 (1993) 97–126
Cetin, B., Burdick, J., Barhen, J.: Global Descent Replaces Gradient Descent to Avoid Local Minima Problem in Learning with Artificial Neural Networks. In: Proc. of the IEEE Conf. on Neural Networks, Vol. 2. (1993) 836–842
Chen, A.M., Lu, H., Nielsen, R.H.: On the Geometry of Feedforward Neural Network Error Surfaces. Neural Computation 5 (1993) 910–927
Gori, M., Tesi A.: On the Problem of Local Minima in Backpropagation. IEEE Trans. on Pattern Analysis and Machine Intelligence 14 (1992) 76–85
Gustafson, K.: Internal Sigmoid Dynamics in Feedforward Neural Networks. Connection Science 10 (1998) 43–73
Hamey, L.G.: XOR Has no Local Minima: A Case Study in Neural Network Error Surface Analysis. Neural Networks 11 (1998) 669–681
Hush, D.R., Horne, B., Salas, J.: Error Surfaces for Multilayer Perceptrons. IEEE Trans. Systems, Man, and Cybernetics 22 (1992) 1152–1161
Kolen, J.F., Pollack, J.B.: Backpropagation is Sensitive to Initial Conditions. Complex Systems 4 (1990) 269–280
Lisboa, P.G., Perantonis, S.J.: Complete Solution of the Local Minima in the XOR Problem. Network 2 (1991) 119–124
Rumelhart, D.E., Hinton, G.E., Williams, R.J.: Learning Internal Representation by Error Propagation. In: Parallel Distr. Proc., Vol. 1. Cambridge, MA: MIT Press (1986) 318–362
Smagt, P.: Minimization Methods for Training Feedforward Neural Networks. Neural Networks 7 (1994) 1–11
Sobol’, I.M., Statnikov, P.B.: Choosing Optimal Parameters in Multicriteria Problems. Nauka, Moscow (1981)
Sontag, E.D., Sussmann, H. J.: Back Propagation Separates Where Perceptrons Do. Neural Networks 4 (1991) 243–249
Sprinkhuizen-Kuyper, I.G., Boers, E.J.: The Error Surface of the Simplest XOR Network Has Only Global Minima. Neural Computation 8 (1996) 1301–1320
Yu, X.H., Chen, G.A.: On the Local Minima Free Condition of Backpropagation Learning. IEEE Trans. Neural Networks 6 (1995) 1300–1303
Author information
Authors and Affiliations
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 1999 Springer-Verlag Berlin Heidelberg
About this paper
Cite this paper
Jordanov, I., Brown, R. (1999). Neural Network Learning Using Low-Discrepancy Sequence. In: Foo, N. (eds) Advanced Topics in Artificial Intelligence. AI 1999. Lecture Notes in Computer Science(), vol 1747. Springer, Berlin, Heidelberg. https://doi.org/10.1007/3-540-46695-9_22
Download citation
DOI: https://doi.org/10.1007/3-540-46695-9_22
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-540-66822-0
Online ISBN: 978-3-540-46695-6
eBook Packages: Springer Book Archive