ヨネダ ケイスケ   Yoneda keisuke
  米田 圭佑
   所属   崇城大学  情報学部 情報学科
   職種   助教
言語種別 英語
発行・発表の年月 2018/12
形態種別 原著論文
査読 査読あり
標題 Distance Metric Learning for The Self-organizing Map using a Co-training Approach
執筆形態 共著
掲載誌名 International Journal of Innovative Computing, Information and Control
掲載区分国外
著者・共著者 Keisuke Yoneda, Tetsuo Furukawa
概要 The aim of this work is to develop a method of distance metric learning for self-organizing maps. We first conducted an investigation in a multi-view learning setting, in which Mahalanobis metrics were determined so that two (or more) views reached a consensus in latent variable estimation. We examined two approaches of multi-view learning: co-training and ensemble. Although both approaches worked as expected, our results suggested that the co-training approach performed better. We further extended the method to a single-view learning setting by introducing the concept of pseudo multi-view learning.