数值计算与计算机应用
數值計算與計算機應用
수치계산여계산궤응용
JOURNAL ON NUMERICAL METHODS AND COMPUTER APPLICATIONS
2000年
2期
112-120
,共9页
李杰星%章云%符曦
李傑星%章雲%符晞
리걸성%장운%부희
noise%neural network%energy function%clustering%robustness%astringency
Allowing for the limitations of LS energy functionused in BP algorithm, this paper proposes a robust learning algorithmbased on the study of how clustering puts down radom noise's effects andthe consideration of intensified training for high-quality examples.Some simulation results demonstrate that the robust algorithm is clearlysuperior to BP algorithm in anti-disturbance and astringency.
