Title :
A tunable Gaussian/square function computation circuit for analog neural networks
Author :
Lin, Shang-Yi ; Huang, Ren-Jiun ; Chiueh, Tzi-Dar
Author_Institution :
Dept. of Electr. Eng., Nat. Taiwan Univ., Taipei, Taiwan
fDate :
3/1/1998 12:00:00 AM
Abstract :
A Gaussian/square function computation circuit suitable for analog neural networks is proposed. It can realize Gaussian and square functions when operating in weak and strong inversion region, respectively. It is shown that the center, width, and peak amplitude of the dc transfer curve can be controlled separably. Measurement results on 3-μm CMOS fabricated chips confirm theoretical and simulation findings
Keywords :
CMOS analogue integrated circuits; analogue processing circuits; circuit tuning; neural chips; 3 micron; CMOS; analog neural networks; dc transfer curve; peak amplitude; simulation; strong inversion region; tunable Gaussian/square function computation circuit; weak inversion region; Analog computers; Circuit optimization; Computer networks; Euclidean distance; MOSFETs; Neural networks; Semiconductor device measurement; Transconductance; Tunable circuits and devices; Very large scale integration;
Journal_Title :
Circuits and Systems II: Analog and Digital Signal Processing, IEEE Transactions on
