Title :
A means of VLSI current controlled weight setting in ANNs
Author :
Singh, S.K. ; Newcomb, R.W. ; Gomez, P. ; Rodellar, V.
Author_Institution :
Dept. of Electr. Eng., Maryland Univ., College Park, MD, USA
Abstract :
A means is given for setting the weights in BiCMOS VLSI realizations of analog neural networks of the Hopfield class. This is achieved by the use of a BJT 4-quadrant current multiplier for which one set of the multiplier´s differential current inputs comes from a CMOS differential pair giving a neuron´s output current from its input capacitor voltage and the other set of the multiplier´s differential currents comes from a differential pair with its input being a weight setting voltage
Keywords :
BiCMOS analogue integrated circuits; Hopfield neural nets; VLSI; multiplying circuits; neural chips; BJT 4-quadrant current multiplier; BiCMOS IC; Hopfield neural nets; VLSI; analog neural networks; current controlled weight setting; input capacitor voltage; multiplier differential currents; Artificial neural networks; Capacitors; Circuits; Equations; Hopfield neural networks; Neurons; Switches; Very large scale integration; Voltage control; Weight control;
Conference_Titel :
Neural Networks, 1995. Proceedings., IEEE International Conference on
Conference_Location :
Perth, WA
Print_ISBN :
0-7803-2768-3
DOI :
10.1109/ICNN.1995.488963
