Supplementary Material for paper "Straightforward Working Principles Behind Modern Data Visualization Approaches"

Citation Author(s):
Jugurta
Montalvão
Federal University of Sergipe (UFS)
Luiz
Miranda
Bernadette
Dorizzi
Submitted by:
Jugurta Montalvao
Last updated:
Tue, 10/06/2020 - 16:41
DOI:
10.21227/jw4g-je18
License:
29 Views
Categories:
0
0 ratings - Please login to submit your rating.

Abstract 

The geometrical reasoning already present in Multidimensional Scaling foundations is directly applied to state-of-the-art visualization algorithms such as t-SNE, LargeVis and UMAP, thus yielding six working principles which are, by hypothesis, enough to produce visual projections qualitatively similar to those obtained with state-of-the-art algorithms. To test and confirm this claim, a simple algorithm is roughly crafted, in which the six working principles are directly implemented, and experiments are done with databases MNIST and Fashion-MNIST. Besides, under the same motivation (of simplification), the problem of visualizing large datasets is tackled through the use of a companion algorithm which is able to project/code new input patterns.

Instructions: 

N.A.