Abstract
This paper analyzes some aspects of the computational power of neural networks using integer weights in a very restricted range. Using limited range integer values opens the road for efficient VLSI implementations because i) a limited range for the weights can be translated into reduced storage requirements and ii) integer computation can be implemented in a more efficient way than the floating point one. The paper concentrates on classification problems and shows that, if the weights are restricted in a drastic way (both range and precision), the existence of a solution is not to be taken for granted anymore. We show that, if the weight range is not chosen carefully, the network will not be able to implement a solution independently on the number of units available on the first hidden layer. The paper presents an existence result which relates the difficulty of the problem as characterized by the minimum distance between patterns of different classes to the weight range necessary to ensure that a solution exists. This result allows us to calculate a weight range for a given category of problems and be confident that the network has the capability to solve the given problems with integer weights in that range.
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
V. Beiu, S. Draghici, and H. E. Makaruk. On limited fan-in optimal neural networks. Technical Report LA-UR-97-2873, Los Alamos National Laboratory, 1997.
D. Brüls. Programmable vlsi array processor for neural networks and matrix-based signal processing. Technical report, Siemens AG, Corporate Research and Development Division, Munich, 1993.
R. Coggins, M. Jabri, and Wattle. A trainable gain analogue vlsi neural network. In Advances in Neural Information Processing Systems, NIPS’93, volume 6, pages 874–881. Morgan Kaufman, 1994.
S. Draghici. On the possibilities of the limited precision weights neural networks in classification problems. In J. Mira, R. Moreno-Diaz, and J. Cabestany, editors, Biological and Artificial Computation: From Neuroscience to Technology, Lecture Notes in Computer Science, pages 753–762. Springer-Verlag, 1997.
S. Draghici. On VLSI-optimal constructive algorithms for classification problems. In Proc. of EIS’98 International Symposium on Engineering of Intelligent Systems, pages 456–462. ICSC Academic Press, 1998.
S. Draghici, V. Beiu, and I. Sethi. A VLSI optimal constructive algorithm for classification problems. In Smart Engineering Systems: Neural Networks, Fuzzy Logic, Data Mining and Evolutionary Programming, pages 141–151. ASME Press, 1997.
G. Dundar and K. Rose. The effect of quantization on multilayer neural networks. IEEE Transactions on Neural Networks, 6:1446–1451, 1995.
M. Duranton. L-neuro 2.3: A vlsi for image processing by neural networks. In Proceedings of the 4-th International Conference on Microelectronics for Neural Networks and Fuzzy Systems, pages 157–160, 1996.
M. A. Glover and W. T. M. III. A massively-parallel simd processor for neural network and machine vision applications. In J. D. Cowan, G. Tesauro, and J. Alspector, editors, Advances in Neural Information Processing Systems, volume 6, pages 843–849. Morgan Kaufmann, 1994.
M. Griffin, G. Tahara, K. Knorpp, and B. Riley. An 11-milion transistor neural network execution engine. In IEEE International Conference on Solid-State Circuits, pages 180–181. IEEE, 1991.
D. Hammerstrom. A highly parallel digital architecture for neural network emulation. In J. G. Delgado-Frias and W. R. Moore, editors, VLSI for Artificial Intelligence and Neural Networks, pages 357–366. Plenum Press, 1991.
M. Höhfeld and S. E. Fahlman. Learning with limited numerical precision using the cascade-correlation algorithm. IEEE Transactions on Neural Networks, 3(4):602–611, 1992.
P. Ienne. Digital connectionist hardware: Current problems and future challenges. In Lecture Notes in Computer Science, Biological and Artificial Computation: From Neuroscience to Technology, volume 1240, pages 688–713. Springer Verlag, 1997.
A. H. Khan and E. L. Hines. Integer-weights neural nets. Electronics Letters, 30(15):1237–1238, 1994.
H. K. Kwan and C. Z. Tang. Designing multilayer feedforward neural networks using simplified activation functions and one-power-of-two weights. Electronics Letters, 28(25):2343–2344, 1992.
H. K. Kwan and C. Z. Tang: Multiplierless multilayer feedfoward neural networks design suitable for continuous input-output mapping. Electronics Letters, 29(14):1259–1260, 1993.
M. Marchesi, G. Orlandi, F. Piazza, L. Pollonara, and A. Uncini. Multilayer perceptrons with discrete weights. In Proceedings International Joint Conference on Neural Networks, volume 2, pages 623–630, 1990.
M. Marchesi, G. Orlandi, F. Piazza, and A. Uncini. Fast neural networks without multipliers. IEEE Transactions on Neural Networks, 4(1):53–62, 1993.
S. Oteki, A. Hashimoto, T. Furuta, T. Watanabe, D. G. Stork, and H. Eguchi. A digital neural network vlsi with on-chip learning using stochastic pulse encoding. In Proceedings of the IJCNN, volume 3, pages 3039–3045, 1993.
U. Ramacher. Synapse—a neurocomputer that synthesizes neural algorithms on a parallel systolic engine. Journal of Parallel and Distributed Computing, 14(3):307–318, March 1992.
C. Z. Tang and H. K. Kwan. Multilayer feedforward neural networks with single power-of-two weights. IEEE Transactions On Signal Processing, 41(8):2724–2727, 1993.
J. B. Theeten, M. Duranton, N. Mauduit, and J. A. Sirat. The lneuro chip: A digital vlsi with on-chip learning mechanism. In International Conference on Neural Networks, pages 593–596, 1990.
J. M. Vincent and D. J. Myers. Weight dithering and wordlength selection for digital backpropagation networks. BT Technology Journal, 10(3):1180–1190, 1992.
M. Viredaz. Design and Analysis of a Systolic Array for Neural Computation. PhD thesis, École Polytechnique Fédérale de Lausanne, 1994.
J. Wawrzynek, K. Asanović, B. Kingsbury, J. Beck, D. Johnson, and N. Morgan. Spert-ii: A vector microprocessor system. Computer, 29(3):79–86, 1996.
Y. Xie and M. A. Jabri. Training algorithms for limited precision feedforward neural networks. Technical Report SEDAL TR 1991-8-3, School of EE, University of Sydney, Australia, 1991.
M. Yasunaga, N. Masuda, M. Yagyu, M. Asai, K. Shibata, M. Ooyama, M. Yamada, T. Sakaguchi, and M. Hashimoto. A self-learning neural network composed of 1152 digital neurons in wafer-scale lsis. In Proceedings of the IJCNN, pages 1844–1849, 1991.
M. Yasunaga, N. Masuda, M. Yagyu, M. Asai, M. Yamada, and A. Masaki. Design, fabrication and evaluation of a 5-inch wafer scale neural network lsi compoosed of 576 digital neurons. In Proceedings of the IJCNN, volume 2, pages 527–536, 1990.
Author information
Authors and Affiliations
Editor information
Rights and permissions
Copyright information
© 1999 Springer-Verlag Berlin Heidelberg
About this paper
Cite this paper
Draghici, S. (1999). On the computational power of limited precision weights neural networks in classification problems: How to calculate the weight range so that a solution will exist. In: Mira, J., Sánchez-Andrés, J.V. (eds) Foundations and Tools for Neural Modeling. IWANN 1999. Lecture Notes in Computer Science, vol 1606. Springer, Berlin, Heidelberg. https://doi.org/10.1007/BFb0098197
Download citation
DOI: https://doi.org/10.1007/BFb0098197
Published:
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-540-66069-9
Online ISBN: 978-3-540-48771-5
eBook Packages: Springer Book Archive