Title :
A current-mode programmable and expandable Hamming neural network
Author :
Li, Guoxing ; Shi, Bingxue
Author_Institution :
Inst. of Microelectron., Tsinghua Univ., Beijing, China
fDate :
6/21/1905 12:00:00 AM
Abstract :
A current-mode programmable and expandable Hamming neural network with the ability to output the first K maximum matching currents from M ones in sorting order is put forward in this paper. The binary template can be programmable or learnable if needed in this network and the K maximum matching currents can be output in sorting order based on the switched-current technique, and its corresponding labels are also output in time sharing mode in the same time. The complexity of this network is just O(N) and its scale can be easily expanded. This network has been fabricated in a 1.2 μm CMOS technology. Both Hspice simulation and experimental results of the prototype chip show good performance
Keywords :
CMOS integrated circuits; VLSI; circuit complexity; current-mode circuits; integrated circuit design; neural chips; programmable circuits; switched current circuits; time-sharing systems; 1.2 mum; Hspice simulation; complexity; current-mode programmable expandable Hamming neural network; learnable binary template; maximum matching currents; programmable binary template; sorting order; switched-current technique; time sharing mode; Artificial neural networks; CMOS technology; Circuits; Microelectronics; Neural networks; Pattern matching; Prototypes; Sorting; Time sharing computer systems; Virtual prototyping;
Conference_Titel :
Neural Networks, 1999. IJCNN '99. International Joint Conference on
Print_ISBN :
0-7803-5529-6
DOI :
10.1109/IJCNN.1999.833450
