Title :
Improved winner-take-all neural network
Author :
Yen, Jui-Chih ; Chang, Silvia
Author_Institution :
Nat. Tsing Hua Univ., Hsinchu, Taiwan
fDate :
3/26/1992 12:00:00 AM
Abstract :
An improved MAXNET is proposed to realise the winner-takes-all function. In IMAX, the strength of mutual inhibitions between each processing unit is adaptively updated. The authors show that IMAX will converge faster than MAXNET. This result is also confirmed by simulations.
Keywords :
neural nets; IMAX; MAXNET; adaptively updated; convergence; mutual inhibitions; mutually inhibitory networks; neural networks; winner-takes-all function;
Journal_Title :
Electronics Letters
DOI :
10.1049/el:19920419
