现代电子技术
現代電子技術
현대전자기술
MODERN ELECTRONICS TECHNIQUE
2014年
8期
12-14,18
,共4页
丁硕%常晓恒%巫庆辉%杨友林%胡庆功
丁碩%常曉恆%巫慶輝%楊友林%鬍慶功
정석%상효항%무경휘%양우림%호경공
Elman神经网络%BP神经网络%模式分类%收敛速度%泛化能力
Elman神經網絡%BP神經網絡%模式分類%收斂速度%汎化能力
Elman신경망락%BP신경망락%모식분류%수렴속도%범화능력
Elman neural network%BP neural network%pattern classification%convergence speed%generalization ability
为了研究Elman神经网络和标准BPNN中何种网络类型更适合于解决模式分类问题,分别构建了基于Elman神经网络的分类模型和基于标准BPNN的分类模型。以平面上二维向量模式的分类为例,对2种分类模型进行训练和泛化能力测试。仿真结果表明,在训练样本数量相等且中小规模网络的条件下,Elman网络模型比BP网络模型具有更高的分类精度,更快的收敛速度,更适合于解决模式分类问题。
為瞭研究Elman神經網絡和標準BPNN中何種網絡類型更適閤于解決模式分類問題,分彆構建瞭基于Elman神經網絡的分類模型和基于標準BPNN的分類模型。以平麵上二維嚮量模式的分類為例,對2種分類模型進行訓練和汎化能力測試。倣真結果錶明,在訓練樣本數量相等且中小規模網絡的條件下,Elman網絡模型比BP網絡模型具有更高的分類精度,更快的收斂速度,更適閤于解決模式分類問題。
위료연구Elman신경망락화표준BPNN중하충망락류형경괄합우해결모식분류문제,분별구건료기우Elman신경망락적분류모형화기우표준BPNN적분류모형。이평면상이유향량모식적분류위례,대2충분류모형진행훈련화범화능력측시。방진결과표명,재훈련양본수량상등차중소규모망락적조건하,Elman망락모형비BP망락모형구유경고적분류정도,경쾌적수렴속도,경괄합우해결모식분류문제。
To study which type of network in Elman neural networks or standard BPNN is more effective for pattern classifi-cation,two classification models based on Elman neural network and standard BPNN are established respectively. The classifica-tion of two- dimensional vector pattern on a plane is taken as an example to train the two classification models and test their generalization abilities respectively. The simulation results show that Elman neural network has higher classification accuracy and faster convergence speed than BPNN under the conditions of the same quantity of the training samples and small or medium size network. And this makes Elman neural network more suitable for solving the problem of pattern classification.