Visualization of multiple class proximity data using proximity embedding and a self-organizing network

Hideaki Misawa, Keiichi Horio, Kazumasa Fukuda, Naoko Ueda, Hatsumi Taniguchi

Research output: Contribution to journalArticlepeer-review


We present a method for visualizing similarity relationships between multiple classes of proximity data. Proximity data consist of dissimilarity or similarity measurements for pairs of objects. Multidimensional scaling (MDS) and its variants can visualize similarity structures of proximity data. However, they cannot visualize similarity relationships between the classes of proximity data. In the presented method, all proximity data are transformed into vectorial data by MDS. After this process, we obtain the class distributions of the proximity data in a vectorial representation. Then, we make a self-organizing map (SOM) of the class distributions by SOM2 algorithm to visualize the similarity relationships between the class distributions. To illustrate the presented method, we apply the method to bacterial flora analysis. We confirmed the possibility that the presented method will be used as an aid for analyzing multiple class proximity data.

Original languageEnglish
Pages (from-to)401-406
Number of pages6
JournalICIC Express Letters, Part B: Applications
Issue number2
Publication statusPublished - Apr 2011
Externally publishedYes

All Science Journal Classification (ASJC) codes

  • General Computer Science


Dive into the research topics of 'Visualization of multiple class proximity data using proximity embedding and a self-organizing network'. Together they form a unique fingerprint.

Cite this