Abstract
Neural network learning is most often understood in the sense of automatic parameter adaptation. Connection strengths between units are typically updated after successive presentations of exemplar data so that the system is able to generalize to previously unseen cases the underlying function or pattern classification rule. Despite the impact of other operational, architectural and analysis aspects, only a minority of the algorithms following this inductive approach focus on parameters others than synaptic weights. In this paper we discuss a pruning method to automatically determine not only the weights but also the topology of a class of learning systems. A procedure to adapt dynamically the pruning strength is also discussed.
Preview
Unable to display preview. Download preview PDF.
References
Chauvin, Y., “Dynamic behavior of constrained back-propagation networks”, in Advances in Neural Information Processing (2), Touretzky, D.S., Ed., pp. 642–649, 1990.
Jacobs, R.A, “Increased rates of convergence through learning rate adaptation”, Neural Networks, vol. 1, pp. 295–307, 1988.
Karnin, E.D., “A simple procedure for pruning back-propagation trained neural networks”, IEEE Trans. Neural Networks, vol. 1, no. 2, pp. 239–242, 1990.
KrishnaKumar, K., “Optimization of the neural net connectivity pattern using a back-propagation algorithm”, Neurocomputing, vol. 5, no. 6, pp. 273–286, 1993.
Kwok, T-Y. and Yeung, D-Y., “Constructive feedforward neural networks for regression problems: A survey”, Tech. Rep. HKUST-CS95-43, The Hong Kong Univ. of Science & Technology, 1995.
Le Cun, Y., Denker, J.S. and Solla, S.A., “Optimal brain damage”, in Advances in Neural Information Processing (2), Touretzky, D.S., Ed., pp. 598–605, 1990.
Mozer, M.C. and Smolensky, P., “Skeletonization: A technique for trimming the fat from a network via relevance assessment”, inAdvances in Neural Information Processing (1), Touretzky, D.S., Ed., pp. 107–115, 1989.
Plaut, D.C., Nowlan, S.J. and Hinton, G.E., “Experiments on learning by back propagation”, Tech. Rep. CMU-CS-86-126, Carnegie Mellon Univ., 1986.
Prechelt, L., “Adaptive parameter pruning in neural networks”, Tech. Rep. 95–009, International Computer Science Institute, Berkeley, CA, 1995.
Reed, R., “Pruning algorithms-A survey”, IEEE Trans. Neural Networks, vol. 4, no. 5, pp. 740–747, 1993.
Riedmiller, M. and Braun, H., “A direct adaptive method for faster backpropagation learning: The RPROP algorithm”, in Proc. of the IEEE International Conference on Neural Networks., pp. 586–591, 1993.
Tollenaere, T., “SuperSAB: Fast adaptive backpropagation with good scaling properties”, Neural Networks, vol. 3, pp. 561–573, 1990.
Weigend, A.S., Rumelhart, D.E. and Huberman, B.A., “Back-propagation, weight-elimination and time series prediction”, in Proc. 1990 Connectionist Models Summer School, Touretzky, D., Elman, J., Sejnowski, T. and Hinton, J., Eds., pp. 105–116, 1990.
Williams, R. and Zipser, D., “A learning algorithm for continually running fully recurrent neural networks”, Neural Computation, vol. 1, no. 2, pp. 270–280, 1989.
Wynne-Jones, M., “Node splitting: A constructive algorithm for feedforward neural networks”, NeuralComputing & Applications, vol. 1 no. 1,pp. 17–22, 1993.
Author information
Authors and Affiliations
Editor information
Rights and permissions
Copyright information
© 1997 Springer-Verlag Berlin Heidelberg
About this paper
Cite this paper
Rementeria, S., Olabe, X. (1997). On simultaneous weight and architecture learning. In: Mira, J., Moreno-Díaz, R., Cabestany, J. (eds) Biological and Artificial Computation: From Neuroscience to Technology. IWANN 1997. Lecture Notes in Computer Science, vol 1240. Springer, Berlin, Heidelberg. https://doi.org/10.1007/BFb0032509
Download citation
DOI: https://doi.org/10.1007/BFb0032509
Published:
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-540-63047-0
Online ISBN: 978-3-540-69074-0
eBook Packages: Springer Book Archive