Intelligent Database Systems Lab

Presenter: YU-TING LU

Authors: Laurens van der Maaten and Geoffrey Hinton

2012. ML

Visualizing non-metric similarities in multiple maps

Intelligent Database Systems Lab

Outlines

MotivationObjectivesMethodologyExperimentsConclusionsComments

Intelligent Database Systems Lab

Motivation• Techniques for multidimensional

scaling(MDS) are subject to the fundamental

limitations of metric spaces in a visualization.

• Multidimensional scaling cannot faithfully

represent intransitive pairwise similarities in

a visualization, and it cannot faithfully

visualize “central” objects.

Intelligent Database Systems Lab

Objectives• This study present an extension of multidimensional

scaling technique multiple maps t-SNE.

• The aims to address the problems of traditional

multidimensional scaling techniques when visualize

non-metric similarities.

• By constructing a collection of maps that reveal

complementary structure in the similarity data.

Intelligent Database Systems Lab

Methodology(review: t-SNE)

Intelligent Database Systems Lab

Methodology-Multiple maps t-SNE

Intelligent Database Systems Lab

ExperimentsResults of multiple maps t-SNE on the word association data set(a-e)

Intelligent Database Systems Lab

ExperimentsResults of multiple maps t-SNE on the word association data set(a-e)

Intelligent Database Systems Lab

ExperimentsResults of multiple maps t-SNE on the word association data set(a-e)

Intelligent Database Systems Lab

ExperimentsResults of multiple maps t-SNE on the word association data set(a-e)

Intelligent Database Systems Lab

ExperimentsResults of multiple maps t-SNE on the word association data set(a-e)

Intelligent Database Systems Lab

ExperimentsResults of multiple maps t-SNE on the word association data set(a-e)

Intelligent Database Systems Lab

Experiments

Intelligent Database Systems Lab

ExperimentsResults of multiple maps t-SNE on the NIPS co-authorship data set(a-d)

Intelligent Database Systems Lab

ExperimentsResults of multiple maps t-SNE on the NIPS co-authorship data set(a-d)

Intelligent Database Systems Lab

ExperimentsResults of multiple maps t-SNE on the NIPS co-authorship data set(a-d)

Intelligent Database Systems Lab

ExperimentsResults of multiple maps t-SNE on the NIPS co-authorship data set(a-d)

Intelligent Database Systems Lab

Experiments

Intelligent Database Systems Lab

Conclusions

• This paper is to construct visualizations that are not

hampered by the two main limitations of metric spaces.

• Apply multiple maps t-SNE to a large data set of word

association data and to a data set of NIPS co-

authorships, demonstrating its ability to successfully

visualize non-metric similarities.

Intelligent Database Systems Lab

Comments

• Advantages- Faithfully visualizing non-metric similarity data

• Applications- Data visualization.- Non-metric similarities.