SOMiMS - Topographic Mapping in the Model Space

• Published in: Intelligent Data Engineering and Automated Learning - IDEAL 2021 Vol. 13113; pp. 502 - 510
• Main Authors: Chen, Xinyue, Shen, Yuan, Zavala, Eder, Tsaneva-Atanasova, Krasimira, Upton, Thomas, Russell, Georgina, Tino, Peter
• Format: Book Chapter
• Language: English
• Published: Switzerland Springer International Publishing AG 2021; Springer International Publishing
• Series: Lecture Notes in Computer Science
• Subjects: Irregular time series; LiMS; Sparse; Topographic mapping
• ISBN: 3030916073; 9783030916077
• ISSN: 0302-9743; 1611-3349
• DOI: 10.1007/978-3-030-91608-4_50

Learning in the model space (LiMS) represents each observational unit (e.g. sparse and irregular time series) with a suitable model of it (point estimate), or a full posterior distribution over models. LiMS approaches take the mechanistic information of how the data is generated into account, thus enhancing the transparency and interpretability of the machine learning tools employed. In this paper we develop a novel topographic mapping in the model space and compare it with an extension of the Generative Topographic Mapping (GTM) to the model space. We demonstrate these two methods on a dataset of measurements taken on subjects in an adrenal steroid hormone deficiency study.