摘要
在统计学习理论中,最近邻法是一种非常直观,重要的学习方法,但是在高维空间中由于"维数灾难"问题,该方法将失效.一种重要的表现形式就是稀疏抽样使得所有样本点都靠近样本的边沿,而不是最近邻样本点.文献[1]中定义的一种中位数距离可以给出它的一个量化的度量,文章从统计的角度对该距离进行了阐述和证明.
In the elements of statistical learning,the nearest-neighbor metbod is a very intuitive and imporant learning method,but the approch and our intuition breaks dowm in high-dimensional space for the＂dimension disaster＂problem. One of the most important manifestations is the sparse sampling makes all the sample points are situated near the edge of the sample, rather than nearest neighbor sample pointsd. The median method in [1] can give it a quantitative measurement. We＇ll expound and proof the distance from the statistical point of view.
出处
《太原师范学院学报（自然科学版）》
2009年第1期1-3,共3页
Journal of Taiyuan Normal University:Natural Science Edition
基金
国家自然科学基金(60873128)
国家社会科学基金青年项目(07CTT022)资助
关键词
高维空间
最近邻法
中位数距离
high-dimensional space
nearest-neighber methods
the median distsnce
