摘要
研究了具有模糊边界样本的网络学习能力，提出了与之相应的ＴｙｐｉｃａｌＩｎｐｕｔＳａｍｐｌｅｓＴｒａｉｎｉｎｇＡｌ－ｇｏｒｉｔｈｍ（ＴＩＳＴＡ）算法并将它与自然时序训练算法进行比较．数学分析和计算机仿真实验结果均证明，ＴＩＳＴＡ算法有效地避免了传统算法的不足，提高了网络的分类性能．
The learning capability of the network with a sample of fuzzy boundary is studied and a typical input sample training algorithm (TISTA) is proposed. The algorithm proposed is compared with the natural sequential training algorithm. It has been proved by both the results of mathematical analysis and computer simulation experiment that the TISTA can overcome the deficiency of the conventional algorithm and improve the classifying capability of the network.
出处
《华中理工大学学报》
CSCD
北大核心
1997年第A01期37-40,共4页
Journal of Huazhong University of Science and Technology
