摘要
本文对照经典的矢量量化算法的不足,讨论了基于竞争机制的连续Hopfield神经网络矢量量化算法的设计与实现。详细描述了网络映射过程、网络能量函数的刻画和神经元状态转换方程。实验结果表明,与经典的LBG算法相比,本文所提算法具有更好的性能和强大的并行处理能力以及更优良的全局优化能力。
This paper discusses the implementation of vector quantization algorithm based on competitive continuous Hopfield neural network contrasting to the defects of traditional VQ algorithm. At the same time,the details on network mapping,energy function constructing and neuro state changing equation have been described. The results of the experiment indicate the performance of the algorithm proposed in this paper is more efficient with powerful parallel a- bility and workable global optimization effect contrasting to LBG.
出处
《计算机科学》
CSCD
北大核心
2004年第9期172-175,共4页
Computer Science
基金
中国科学院知识创新工程方向性研究项目基金(名称:大型数字对象应用环境及其并行模拟
批准号:KGCX2-JG-09)
