摘要
对自组织特征映射神经网络的特性进行分析，并将其与矢量量化问题的实质进行比较，提出了一个实现矢量量化的自组织特征映射算法。分析与实验表明，该算法是稳定收敛的。算法的学习结果与网络的初始状态无关，并且十分接近于全局最优解的下限。将该算法应用于图像数据的压缩。
The characteristics of SOFM neural network is analysed and compared with the feature of Vector Quantizing problem in this paper. Based on this an algorithm for Vector Quantizing is put forward. Analysis and experiments show that this algorithm is stably convergent. The study results are irrelevant to the initial status of the network and quite approximate to the lower limit of the optimum solutions. Very satisfied results are achieved when this algorithm is used for graphic data compression.
出处
《南京理工大学学报》
EI
CAS
CSCD
1999年第5期393-396,共4页
Journal of Nanjing University of Science and Technology
