Applications of Average Geodesic Distance in Manifold Learning

摘要：

Manifold learning has become a hot issue in the research fields of machine learning and data mining.Current manifold learning algorithms assume that the observed data set has the high density.But,how to evaluate the denseness of the high dimensional observed data set? This paper proposes an algorithm based on the average geodesic distance as the preprocessing step of manifold learning.Moreover,for a high dense data set evaluated,we further utilize the average geodesic distance to quantitatively analyze the mapping relationship between the highdimensional manifold and the corresponding intrinsic low-dimensional manifold in the known ISOMAP algorithm.Finally,experimental results on two synthetic Swiss-roll data sets show that our method is feasible.

作者: Xianhua Zeng

作者单位: School of Computer Science,China West Normal University

会议类型: 国际会议

会议名称: The Third International Conference on Rough Sets and Knowledge Tevhnology(RSKT 2008)(第三届粗糙集与知识技术国际会议)

会议地点: 成都

会议语种:英文

页码: 540-547

在线出版日期: 2008-05-17（万方平台首次上网日期，不代表论文的发表时间）