Abstract
This work is concerned with the establishment of a relation between the fields of qualitative reasoning (QR) and neural networks. We explore how well-known backpropagation learning algorithm can be studied from the point of view of QR Qualitative models are based on the discretization of their parameters and the use of closed operators on the sets induced by the discretization. Henceforth, a qualitative version of backpropagation is an algorithm in which the variables involved in it belong to one of the finite classes defined. We analyse the algorithms resulting from this transformation and test their performance with a set of four problems. The results are encouraging and provide an empirical basis for a deeper, theoretical study, which can be very useful to realize physical implementations of the algorithm or as a starting point for the development of reinforcement learning algorithms.
This is a preview of subscription content, log in via an institution to check access.
Preview
Unable to display preview. Download preview PDF.
Bibliography
Jeffrey L. Elman. Finding structure in time. CRL Technical Report 8801, 1988.
Barto A.G., Sutton R.S., and Brower. Associative Search Network; A Reinforcement Learning Associative Memory. Biological Cybernetics, 40: 201–211, 1981.
Barto A.G., Sutton R.S., Anderson C.W. Neuronlike elements that can solve difficult learning control problems. IEEE Trans. on Systems, Man, and Cybernetics, 13: 835–846, 1983.
Carrabina, J. Xarxes Neuronals VLSI d'alta velocitat. Doctoral dissertation. Universität Autònoma de Barcelona, 1991.
Sivilotti M.A., Emerling M.R., Mead C.A. VLSI Architectures for Implementation of Neural Networks. American Institute of Physics conference Proceedings 151, Neural Networks for Computing, pp 408–413, 1986.
Pérez Vicente C., Learning algorithm for binary synazpsis. Statistical Mechanics of Neural Networks, Springer Verlag 1990.
Forbus K.D., Qualitative Process Theory, Artificial Intelligence 24, 1984.
Rumelhart D.E., Hinton G.E., Williams R.J. Learning internal representations by error propagation. Parallel Distributed Processing: Exploration in the Microstructures of Cognition, vol.1, pp 318–362, Mit Press, 1986.
Buntine, W.L. and Weigend, A.S., Computing Second Order Derivatives in Feed-Forward Networks: a Review. IEEE Transactions on Neural Networks, (revised February 1993).
Author information
Authors and Affiliations
Editor information
Rights and permissions
Copyright information
© 1995 Springer-Verlag Berlin Heidelberg
About this paper
Cite this paper
Seix, B.M., Mallofré, A.C., Carreté, N.P. (1995). Qualitative approach to gradient based learning algorithms. In: Mira, J., Sandoval, F. (eds) From Natural to Artificial Neural Computation. IWANN 1995. Lecture Notes in Computer Science, vol 930. Springer, Berlin, Heidelberg. https://doi.org/10.1007/3-540-59497-3_212
Download citation
DOI: https://doi.org/10.1007/3-540-59497-3_212
Published:
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-540-59497-0
Online ISBN: 978-3-540-49288-7
eBook Packages: Springer Book Archive