报告题目：Nonmetric Multidimensional Scaling: Feasibility, Algorithms and Applications
主 讲 人：李 庆 娜
腾 讯 ID：648 168 751
Nonmetric multidimensional scaling(NMDS) is an important tool in data science to deal with dissimilarity data. In this talk, we will discuss the feasibility, numerical algorithms and the applications of NMDS, mainly based on the rank constraint Euclidean distance matrix model for NMDS. Despite the long history of NMSD, the feasibility issue of NMDS has been rarely discussed, which motivates us to take a systematical investigation of it. The challenges of designing efficient numerical algorithms for NMDS are the nonconvex constraint as well as the huge number of ordinal constraints. We will also discuss several numerical algorithms for NMDS, trying to tackling the two challenges in different ways. For applications, besides the traditional application such as sensor network localization, protein molecular conformation, we will also apply NMDS model to image ranking and posture sensing.