Title :
Design of modifiable-weight synapse CMOS analog cell
Author :
Ibrahim, Fawzy ; Zaghloul, M.E.
Author_Institution :
Dept. of Electr. Eng. & Comput. Sci., George Washington Univ., Washington, DC, USA
Abstract :
The design of a CMOS analog VLSI synapse circuit that learns according to supervised or unsupervised learning rules is introduced. The sign of the weight change is determined by using a weight-change sensor circuit, and then a definite weight-change value is added to the old weight using a weight-updating circuit. The resulting system provides a small weight change, positive and negative weight values, and a simple controlling weight-change mechanism. Simulation results based on the Analog Workbench software are described
Keywords :
CMOS integrated circuits; VLSI; learning systems; linear integrated circuits; neural nets; Analog Workbench software; controlling weight-change mechanism; modifiable-weight synapse CMOS analog cell; negative weight values; supervised learning; unsupervised learning; weight-change sensor circuit; weight-change value; weight-updating circuit; Biosensors; CMOS analog integrated circuits; CMOS technology; Circuit simulation; Clocks; Neurons; Parallel processing; Unsupervised learning; Very large scale integration; Weight control;
Conference_Titel :
Circuits and Systems, 1990., IEEE International Symposium on
Conference_Location :
New Orleans, LA
DOI :
10.1109/ISCAS.1990.112636
