Title :
Modular neural network hardware implementation of analog-digital mixed operation
Author_Institution :
Korea Telecom Res. Center, Seoul, South Korea
Abstract :
General-purpose neural network hardware and digital computer interfaces are described. A custom-designed CMOS neurochip is employed to construct, with a chip performance of 2 G-connections/s in speed, thousands of connections in size, and 8-b weight accuracy. A basic hardware module consisting of several chips and digital computer interfaces yields 10 G-connections/s in speed and more than tens of thousands of connections is size
Keywords :
CMOS integrated circuits; VLSI; mixed analogue-digital integrated circuits; multichip modules; neural chips; CMOS neurochip; analog-digital mixed operation; modular neural network; module; Analog-digital conversion; CMOS technology; Computer interfaces; Electric resistance; MOSFET circuits; Neural network hardware; Neural networks; Switches; Very large scale integration; Voltage;
Conference_Titel :
Neural Networks, 1993., IEEE International Conference on
Conference_Location :
San Francisco, CA
Print_ISBN :
0-7803-0999-5
DOI :
10.1109/ICNN.1993.298641
