DocumentCode :
972573
Title :
Multilayer parallel distributed pattern recognition system model using sparse RAM nets
Author :
Guoqing, Yang ; Songcan, Chen ; Jun, Lu
Author_Institution :
Dept. of Comput. Sci. & Eng., Nanjing Aeronaut. Inst., China
Volume :
139
Issue :
2
fYear :
1992
fDate :
3/1/1992 12:00:00 AM
Firstpage :
144
Lastpage :
146
Abstract :
The authors propose a novel multilayer parallel distributed pattern recognition system model in which the n-tuple principle in WISARD is followed and ordinary RAM nets are replaced by sparse RAM. The new system considers pattern correlation and is able to optimise n-tuple size a larger range through reduction of cost. Preliminary experiments with handwritten Chinese character recognition have confirmed the feasibility of the model.
Keywords :
character recognition; content-addressable storage; neural nets; random-access storage; WISARD; associative memory; cost reduction; handwritten Chinese character recognition; multilayer parallel distributed pattern recognition system; n-tuple principle; n-tuple size; pattern correlation; sparse RAM nets;
fLanguage :
English
Journal_Title :
Computers and Digital Techniques, IEE Proceedings E
Publisher :
iet
ISSN :
0143-7062
Type :
jour
Filename :
129254
Link To Document :
بازگشت