Abstract
This paper describes a modular analog VLSI architecture for the implementation of artificial neural networks. Analog neural network implementations are faster and smaller than their digital counterparts, but the problem of smaller dynamic range of the analog weight memory and the linearity of the synapses based on analog multipliers increases the need for design effort at the circuit level. We suggest that a complex neural network system can be implemented in a single chip if a modular architecture design using simple analog circuits is followed. To demonstrate the VLSI implementability of the neural network system, a description of each analog circuit block is provided.
Preview
Unable to display preview. Download preview PDF.
Similar content being viewed by others
Bibliography
P.E. Allen and D. R. Holberg, CMOS Analog Circuit Design, New-York, NY, Holt, Rinehart and Winston, Inc., 1987.
L.N.M. Edward, “Comment on “Voltage-Controlled Linear Resistor by Two MOS Transistors and its Application to Active RC Filter MOS Integration”, in Proc. IEEE, Vol. 74, No. 5, pp. 753–755, 1986.
R.L. Geiger, P. E. Allen, and N. R. Strader, VLSI Design Techniques for Analog and Digital Circuits, New-York, NY, McGraw-Hill, 1990.
P.R. Gray and R. G. Meyer, Analysis and Design of Analog Integrated Circuits, 3rd ed., New-York, NY, John Wiley & Sons, 1993.
K.K. Moon, F.J. Kub and I.A. Mack, “Random Address 32×32 Programmable Analog Vector-Matrix Multiplier for Artificial Neural Networks”, in Proc. IEEE CICC '90, pp. 26.7.1–26.7.4, 1990.
V. Radeka, “Fast Analogue Multipliers With Field-Effect Transistors”, IEEE Trans. Nuclear Science, Vol. 11, No. 1, pp. 302–307, 1964.
F.M.A. Salam and M. R. Choi, “Analog MOS Vector Multipliers for the Implementation of Synapses in Artificial Neural Networks”, Journal of Circuits, Systems, and Computers, Vol. 1, No. 2, pp. 205–228, 1991.
S. Satyanarayana, Y. P. Tsividis, and H. P. Graf, “A Reconfigurable VLSI Neural Network”, IEEE J. Solid-State Circuits, Vol. SC-27, No. 1, pp. 67–81, 1992.
O. Vermesan, The MOS Transistor as the Basic Building Block for Analog VLSI Implementation of Neural Networks. Scientific/Technical Report No. 1994-11, ISSN 0803-2696, University of Bergen, Norway, 1994.
O. Vermesan, Memory Units for Analog VLSI Implementation of Neural Networks. Scientific/Technical Report No. 1994-12, ISSN 0803-2696, University of Bergen, Norway, 1994.
O. Vermesan, Neural Networks Implementation — Issues and Techniques. Scientific/Technical Report No. 1994-20, ISSN 0803-2696, University of Bergen, Norway, 1994.
O. Vermesan, and A.I. Vermesan, “The Use of Hybrid Intelligent Systems in Telecommunications” In J. Liebowitz and D.S. Prerau, Eds., Worldwide Intelligent Systems-Approaches to Telecommunications and Network Management, Amsterdam, IOS Press, Chapter 10, pp. 186–226, 1995.
Author information
Authors and Affiliations
Editor information
Rights and permissions
Copyright information
© 1995 Springer-Verlag Berlin Heidelberg
About this paper
Cite this paper
Vermesan, O. (1995). A modular VLSI architecture for neural networks implementation. 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_252
Download citation
DOI: https://doi.org/10.1007/3-540-59497-3_252
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