摘要
对当前具有代表性的几种特征提取算法进行了分析与比较,并在Bourgain算法的基础上,提出一种基于数据类别数及各类代表元素等启发式信息的复杂数据特征提取算法。对于M类复杂数据,该算法可以提取出维向量用来表示这些数据。针对实际数据,对几种算法的降维性能进行了比较实验,实验结果表明该算法具有很好的特征提取效果。
This paper analyzes several existing feature extraction algorithms. Based on Bourgain, the paper proposes an improved algorithm, it uses the heuristics coming from the data itself in selecting reference sets and reference elements. It can reduce the dimensions of the data of M categories to dimensions. The paper has done some experiments on some practical document data to compare the performance of those algorithms .The experimental results show that the algorithm outperforms them in feature extraction.
出处
《计算机工程》
CAS
CSCD
北大核心
2003年第19期68-69,179,共3页
Computer Engineering
基金
国家自然科学基金资助项目(60005004)
安徽省自然科学基金资助项目(01042302)
关键词
高维数据
特征提取
降维
High-dimensional data
Feature extraction
Dimension reduction
