内蒙古师范大学学报(自然科学汉文版)
內矇古師範大學學報(自然科學漢文版)
내몽고사범대학학보(자연과학한문판)
JOURNAL OF INNER MONGOLIA NORMAL UNIVERSITY(NATURAL SCIENCE EDITION)
2015年
4期
520-522
,共3页
BP 神经网络%样本空间%归一化%主成分分析%聚类%神经网络集成
BP 神經網絡%樣本空間%歸一化%主成分分析%聚類%神經網絡集成
BP 신경망락%양본공간%귀일화%주성분분석%취류%신경망락집성
BP neural network%sample space%uniformization%PCA%clustering
泛化能力是评价 BP 神经网络的重要指标之一,没有良好的泛化能力,网络就不能够应用和推广。网络样本质量是决定 BP 神经网络能否训练成功的关键要素。对目前常用的几种样本处理方法进行总结和归纳,比较了每种处理方法的优劣。
汎化能力是評價 BP 神經網絡的重要指標之一,沒有良好的汎化能力,網絡就不能夠應用和推廣。網絡樣本質量是決定 BP 神經網絡能否訓練成功的關鍵要素。對目前常用的幾種樣本處理方法進行總結和歸納,比較瞭每種處理方法的優劣。
범화능력시평개 BP 신경망락적중요지표지일,몰유량호적범화능력,망락취불능구응용화추엄。망락양본질량시결정 BP 신경망락능부훈련성공적관건요소。대목전상용적궤충양본처리방법진행총결화귀납,비교료매충처리방법적우렬。
Generalization ability is one important index for BP neural network.The network cannot implemented or promoted without a good generalization ability.Among numerous factors affecting the BP neural network generalization ability,network sample quality is a key to determining the success of the BP neural network training.In this paper,a number of commonly used sample processing methods are summa-rized and classified.Comparisons are made among various approaches in terms of their advantages and disadvantages.Future research has been proposed in the sample processing methods.