Efficient sigmoid function for neural networks based FPGA design

Authors:
Xi Chen;Gaofeng Wang;Wei Zhou;Sheng Chang;Shilei Sun
Affiliations:
School of Electronic Information, Wuhan University, Wuhan, Hubei, China;School of Electronic Information, Wuhan University, Wuhan, Hubei, China;School of Electronic Information, Wuhan University, Wuhan, Hubei, China;Institute of C.J. Huang Information Technology, Wuhan University, Wuhan, Hubei, China;Institute of C.J. Huang Information Technology, Wuhan University, Wuhan, Hubei, China
Venue:
ICIC'06 Proceedings of the 2006 international conference on Intelligent Computing - Volume Part I
Year:
2006

Citing 3
Cited 0

Multilayer feedforward networks are universal approximators

Neural Networks
Hybrid CORDIC Algorithms

IEEE Transactions on Computers
A unified algorithm for elementary functions

AFIPS '71 (Spring) Proceedings of the May 18-20, 1971, spring joint computer conference

Quantified Score

Hi-index	0.00

Visualization

Abstract

Efficient design of sigmoid function for neural networks based FPGA is presented. Employing the hybrid CORDIC algorithm, the sigmoid function is described with VHDL in register transfer level. In order to enhance the efficiency and accuracy of implementation on Altera’s FPGA, the technology of pipeline and look-up table have been utilized. Through comparing the results obtained by the post-simulation of EDA tools with the results directly accounted by Matlab, it can be concluded that the designed model works accurately and efficiently.