特征主元提取与自组织影射的剖析

The Analysis of Feature Principal Component Extraction and Self organizing Map

剖析了用神经网络实现特征主元提取（ＰＣＥ）、自组织特征影射（ＳＯＦＭ）、类扩展自组织语义影射（ＳＯＳＭ）和改进的特征细化自组织影射．通过对运载工具的特征压缩，进行可视性分析，结果表明ＰＣＥ和ＳＯＦＭ都能显示事物间的类似程度和关系结构，具有语义影射的功能．特征细化的ＳＯＦＭ同样能达到类扩展ＳＯＳＭ细化分类的功能，它克服了类扩展的ＳＯＳＭ增加输入特征的维数、增加不必要的计算量、输入特征与影射结果不相一致的缺点．

Analyses the neural networks of the feature principal comonent extraction(PCE),the self organizing feature map(SOFM),the classes augment self organizing semantic map(SOSM)and improved feature fine quantization self organizing map.By means of the feature compression of vehicles and vision analysis,the result indicates that PCE and SOFM can show similarity between objects and relative structures,have function of semantic map.The SOFM of feature fine quantization can achieve detail classfication as classes augment SOSM, it overcomes the drawbacks of increasing dimensions of SOSM augments input feature,unnecessary calculation and inconsistency of input feature and map result.