摘要
本文针对基于自组织映射的后非线性独立分量分析方法的缺点,提出了一种具有全局拓扑保持特性的网络权值初始化方法,该方法不仅提高了网络的收敛速度,而且有效地避免了网络陷入局部极小;同时,在混合方式相同的情况下,可使分离信号的次序和符号保持不变.为了检验该方法的拓扑保持特性,本文还提出了一个简单的拓扑度量函数.通过仿真实验证实了所提出的方法是有效的.
According to the drawbacks of post-nonlinear Independent Component Analysis (ICA) method based on Self-Organizing Maps (SOM), an initialization method with global topology preservation property for SOM is proposed. By using this method, not only the convergence speed of SOM is improved greatly, but also the probability of the network falling into local minima is reduced effectively. Furthermore, when the mixed manner is fixed, the order and the sign of the estimated sources are invariable. In order to verify the topology preservation property of our method, a simple topographic function is proposed. The simulation results show that the proposed method is effective.
出处
《电子学报》
EI
CAS
CSCD
北大核心
2005年第5期889-892,共4页
Acta Electronica Sinica
基金
国家自然科学基金(No.60472103)
关键词
自组织映射
后非线性混合
独立分量分析
盲源分离
Computer simulation
Convergence of numerical methods
Functions
Independent component analysis
Probability
Self organizing maps
Topology