期刊
PATTERN RECOGNITION LETTERS
卷 164, 期 -, 页码 148-152出版社
ELSEVIER
DOI: 10.1016/j.patrec.2022.11.007
关键词
-
Conditional multidimensional scaling is a method that seeks for a low-dimensional configuration from pairwise dissimilarities in the presence of other known features. It simplifies the knowledge discovery process and has broad applications. This paper proposes an alternative closed-form solution based on multiple linear regression and eigendecomposition to address the limitations of the current method.
Conditional multidimensional scaling seeks for a low-dimensional configuration from pairwise dissimilarities, in the presence of other known features. This method enables a simpler knowledge discovery process. Thus, it has broad application across different science and engineering domains because prior information of such known features is often available. The current solution of conditional multidimensional scaling is obtained via minimizing its conditional stress objective function with conditional SMACOF, an iterative optimization algorithm. However, iterative optimization is sensitive to starting values and can be time consuming for large problems. This paper proposes an alternative closed-form solution for conditional multidimensional scaling to address these deficits. The proposed method is based on multiple linear regression and eigendecomposition. The proposed algorithm does not necessarily replace conditional SMACOF. The former can be used to initialize the latter to improve its speed and accuracy. (c) 2022 Elsevier B.V. All rights reserved.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据