Title :
Reduced power SOM/LVQ arrays through distance thresholding
Author :
Hendry, D.C. ; Cambio, R.
Author_Institution :
Dept. of Eng., Univ. of Aberdeen, UK
Abstract :
The distance thresholding method is introduced as a means of reducing power dissipation for self-organising map (SOM)/learning vector quantisation (LVQ) calculations using a single instruction multiple data (SIMD) array. The method requires little additional hardware support. For three standard SOM/LVQ benchmarks power reductions are approximately 45%.
Keywords :
learning (artificial intelligence); parallel processing; self-organising feature maps; vector quantisation; SIMD array; SOM/learning vector quantisation; clock gating technique; distance thresholding method; power dissipation reduction; reduced power SOM/LVO arrays; self-organising map; single instruction multiple data array;
Journal_Title :
Electronics Letters
DOI :
10.1049/el:20030961
