摘要
综合了标准BP算法与“批处理”BP算法的各自特点,提出了一种新的BP网络的学习算法.该算法既具有“批处理”BP算法收敛时迭代次数少的优点,又能克服“批处理”算法对大样本集进行学习时每次计算量较大且收敛时间长的缺点.该算法具有“批处理”算法同阶的迭代次数,但每次迭代所需计算工作量大约是“批处理”
To synthesize the advantages of standard BP algorithm and “batch learning”BP algorithm,A new algorithm is put forword.It not only has the advantage of fewer iteration times like the “batch learning”algorithm,but also overcomes the disadvantages of much computational work and long learning time while learning big samples.The examples show that this algorithm has the same iteration times as the “batch learning”algorithm,but its computational work of every iteration is much less than that for the “batch learning”algorithm.
出处
《河海大学学报（自然科学版）》
CAS
CSCD
2000年第1期113-115,共3页
Journal of Hohai University(Natural Sciences)
