DUGi

Rate-Distortion Theory for Clustering in the Perceptual Space

http://dugi.udg.edu/item/http:@@@@hdl.handle.net@@2072@@292125

How to extract relevant information from large data sets has become a main challengein data visualization. Clustering techniques that classify data into groups according to similaritymetrics are a suitable strategy to tackle this problem. Generally, these techniques are applied in thedata space as an independent step previous to visualization. In this paper, we propose clusteringon the perceptual space by maximizing the mutual information between the original data and thefinal visualization. With this purpose, we present a new information-theoretic framework based onthe rate-distortion theory that allows us to achieve a maximally compressed data with a minimalsignal distortion. Using this framework, we propose a methodology to design a visualization processthat minimizes the information loss during the clustering process. Three application examples of theproposed methodology in different visualization techniques such as scatterplot, parallel coordinates,and summary trees are presented

This work has been funded in part by grants from the Spanish Government (Nr. TIN2016-75866-C3-3-R) and from the Catalan Government (Nr. 2014-SGR-1232)

MDPI (Multidisciplinary Digital Publishing Institute)