MULTIPLIERLESS MULTILAYER FEEDFORWARD NEURAL-NETWORK DESIGN SUITABLE FOR CONTINUOUS INPUT-OUTPUT MAPPING

被引：9
作者：
KWAN, HK
TANG, CZ
机构：
[1] Department of Electrical Engineering, University of Windsor, Windsor, Ontario N9B 3P4
来源：
ELECTRONICS LETTERS | 1993年 / 29卷 / 14期
关键词：
NEURAL NETWORKS;
D O I：
10.1049/el:19930841
中图分类号：
TM [电工技术]; TN [电子技术、通信技术];
学科分类号：
0808 ; 0809 ;
摘要：
A method for designing a multiplierless multilayer feed-forward neural network (MFNN) suitable for continuous input-output mapping is presented. When tested with noisy vectors, the network can retain a similar recall accuracy as the corresponding MFNN with continuous weights. The advantages of the design method include faster computational speed and reduced digital hardware cost.
引用
收藏
页码：1259 / 1260
页数：2
相关论文
共 7 条
[1]   SYSTOLIC ARRAY ALGORITHM FOR THE HOPFIELD NEURAL-NETWORK GUARANTEEING CONVERGENCE [J].
EUN, S ;
KIM, JS ;
MAENG, SR ;
YOON, H .
ELECTRONICS LETTERS, 1993, 29 (07) :609-611
[2]   REGISTER REQUIREMENTS FOR LINEAR-ARRAY NEURAL NETWORK EMULATORS [J].
HUNTER, J ;
JONES, S .
ELECTRONICS LETTERS, 1992, 28 (19) :1804-1805
[3]  
KUNG SY, 1988, 2ND P IEEE INT C NEU, V2, P165
[4]   SIMPLE SIGMOID-LIKE ACTIVATION FUNCTION SUITABLE FOR DIGITAL HARDWARE IMPLEMENTATION [J].
KWAN, HK .
ELECTRONICS LETTERS, 1992, 28 (15) :1379-1380
[5]   DESIGNING MULTILAYER FEEDFORWARD NEURAL NETWORKS USING SIMPLIFIED SIGMOID ACTIVATION FUNCTIONS AND ONE-POWERS-OF-2 WEIGHTS [J].
KWAN, HK ;
TANG, CZ .
ELECTRONICS LETTERS, 1992, 28 (25) :2343-2345
[6]   METHOD FOR GENERATING PARTIALLY CORRELATED VECTOR SETS FOR NEURAL-NETWORK SIMULATIONS [J].
KWAN, HK ;
WANG, ZW ;
SOLTIS, JJ .
INTERNATIONAL JOURNAL OF ELECTRONICS, 1993, 74 (04) :523-528
[7]  
KWAN HK, 1989, MAY P IEEE INT S CIR, V2, P790
1