On the Reduction of the Number of Coefficient Circuits in a DTCNN Cell

This paper introduces a methodology to reduce the number of coefficient circuits in a DTCNN cell without penalty at application level. Trade-offs like area-processing time, and some other figures of merit like accuracy and power dissipation are considered. It is shown that it is possible to obtain efficient implementations with a reduced number of coefficient circuits. Some examples illustrate the proposal

[1]  P.R. Kinget.  Device mismatch and tradeoffs in the design of analog circuits , 2005, IEEE Journal of Solid-State Circuits.

[2]  Kari Halonen,et al.  An analog array processor hardware realization with multiple new features , 2002, Proceedings of the 2002 International Joint Conference on Neural Networks. IJCNN'02 (Cat. No.02CH37290).

[3]  Kari Halonen,et al.  High density VLSI implementation of a bipolar CNN with reduced programmability , 2004, 2004 IEEE International Symposium on Circuits and Systems (IEEE Cat. No.04CH37512).

[4]  Ari Paasio,et al.  A 176 x 144 processor binary I/O CNN-UM chip design , 1999 .

[5]  Victor M. Brea,et al.  A one-quadrant discrete-time cellular neural network architecture for pixel-level snakes: B/W processing , 2005, 2005 IEEE International Symposium on Circuits and Systems.

[6]  D. Cabello,et al.  On the emulation of large-neighborhood templates with binary CNN-based architectures , 2005, 2005 9th International Workshop on Cellular Neural Networks and Their Applications.

[7]  Csaba Rekeczky,et al.  Implementation of a pixel-level snake algorithm on a CNNUM-based chip set architecture , 2004, IEEE Transactions on Circuits and Systems I: Regular Papers.

[8]  V. Bonaiuto,et al.  Time division digital programmable OTA for cellular neural networks , 2005, Proceedings of the 2005 European Conference on Circuit Theory and Design, 2005..