Color-Based Model for Dimensionality Reduction

Diego Fernando Peña Unigarro, Universidad de Nariño, San Juan de Pasto - Colombia 2016

Nowadays, a consequence of data overload is that 
world’s technology capacity to collect, communicate, and 
store large volumes of data is increasing faster than hu
man analysis skills. Such an issue has motivated the de
velopment of graphic ways to visually represent and ana 
lyze high-dimensional data. 

Figure 1. Embedded space from Coil-20 data-set
Figure 1. Embedded space from Coil-20 data-set.

Particularly, in this work, we 
propose a graphical interface that allow the combination of 
dimensionality reduction (DR) methods using a chromatic 
model to make data visualization more intelligible for hu
mans. 

Chromatic model
Figure 2. Chromatic model and The weighting factors.




Graphic interface

The interface is designed for an easy and interactive 
use, so that input parameters are given by the user via the 
selection of RGB values inside a given surface. Proposed 
interface enables (even non-expert) users to intuitively ei
ther select a concrete DR method or carry out a mixture of 
methods. 

Figure 3. Graphical interface with all its functions

Tutorial Video

Publications

Thesis


Papers

SYMPOSIUM ON SIGNAL PROCESSING, IMAGES AND ARTIFICIAL VISION (STSIVA)


- Interactive Visualization Methodology of High-Dimensional Data with a 
Color-Based Model for Dimensionality Reduction


IBEROAMERICAN CONGRESS ON PATTERN RECOGNITION (CIARP)

- Interactive data visualization using dimensionality 
reduction and similarity-based representations

Personal Information

                
DIEGO FERNANDO PEÑA UNIGARRO 
Electronic engineer from the University of                   Nariño in Pasto - Colombia.
 Email
diferpun@gmail.com
diferpun@hotmail.com

Comments