摘要
提出了一种基于可靠稳定的模糊核学习矢量量化(FKLVQ)聚类的Sammon非线性映射新算法。该方法通过Mercer核,将数据空间映射到高维特征空间,并在此特征空间上进行FKLVQ学习获取数据空间有效且稳定的聚类权矢量,然后在特征空间和输出空间上仅针对各空间的数据样本和它们各自的聚类权矢量进行Sammon非线性核映射。这样既降低了计算的复杂度,又使数据空间和输出空间上数据点与聚类中心间的距离信息保持相似。仿真结果验证了该方法的可靠性和稳定性。
An new algorithm for Sammon's nonlinear kernel mapping based on reliable and stable fuzzy kernel learning vector quantization was presented. The data space was mapped to high dimension feature space with Mercer kernel function, and fuzzy kernel learning vector quantization (FKLVQ) was done on the feature space to obtain the effective and stable clustering weight vectors. Finally Sammon's nonlinear kernel mapping only for the data points and the clusters was executed on the output space and the feature space, thus reducing computational complexity and preserving the distance resemblance between the clusters and the data points from the data space to the output space. Simulation results demonstrate the reliability and stability of the proposed algorithm.
出处
《计算机应用》
CSCD
北大核心
2007年第3期553-555,共3页
journal of Computer Applications
基金
国家自然科学基金资助项目(60172011)
关键词
非线性映射
Sammon投影
距离保持性
计算复杂度
模糊核
学习矢量量化
nonlinear mapping
Sammon's projection
distance preservation
computational complexity
fuzzy kernel
Learning Vector Quantization (LVQ)