摘要
Sammon映射算法将特征数据从高维映射到低维可视空间,并保持了高维空间中数据样本点之间的距离,可以对数据特征的有效性进行直观的可视化研究。利用自组织映射对特征数据的样本数量进行压缩预处理,降低Sammon算法的计算量,由此提出了改进型的SOSammon算法。通过对实测数据的分析表明,改进算法速度上优于原始算法,能够较好显示个体特征的散布特性。
Sammon mapping is an non-linear projection to mensional image, well keeping the dissimilarity of original visualize high-dimensional data as low-didata points, it can be applied as analytic technique of emitter individual feature. It is presented that a new Sammon algorithm named SOSammon using self-organization mapping to compress the pattern number of feature data, which reduce the large numerical calculation. The analysis result of real data shows that the improved algorithm has better performance, and has better vision of dispersion characteristic of individual feature.
出处
《电子信息对抗技术》
2010年第1期21-24,40,共5页
Electronic Information Warfare Technology
关键词
有效性分析
可视化
Sammon映射
降维
个体识别
validity analysis
visualization
Sammon mapping
dimensional reduction
individual identification
