Connectionists: ESANN 2015 special session and contest about "Unsupervised Nonlinear Dimensionality Reduction"

[John Lee]

Apologies for cross-posting

ESANN 2015 special session and contest

"Unsupervised Nonlinear Dimensionality Reduction"

Call for papers and contest participation

---

About ESANN 2015 (April 22-24, Bruges, Belgium):

http://www.esann.org/
https://www.elen.ucl.ac.be/esann/index.php?pg=specsess

More info about the contest and special session:

https://sites.google.com/site/nldrcontest/

---

Organizers:

John A. Lee (Université catholique de Louvain, Belgium)
Kerstin Bunte (Aalto University, Finland)

---

Summary:

Nonlinear dimensionality reduction (NLDR) is a long-standing problem
that has motivated the development of many different techniques,
starting from early nonlinear variants of multidimensional scaling.
These techniques have to face major difficulties related to the curse of 
dimensionality,
like norm concentration and hubness. In order to tackle the NLDR challenge,
several paradigms have been investigated, such as linear algebra, graph 
theory,
reproducing kernel Hilbert spaces, neural networks, etc.

This special session of ESANN 2015 welcomes contributions to the field 
of NLDR,
in the form of novel techniques and methods, new quality assessment tools,
or variants of existing methods with specific applications,
like big data and/or mixed data (quantitative, categorical, etc.).
Among possible topics:

     Unsupervised/semi-supervised/supervised NLDR
     Manifold learning
     Parametric/non-parametric NLDR models
     Out-of-sample extensions
     Specific data structures for accelerated NLDR
     Real-life applications of NLDR in industry, environment science, etc.

The submitted papers will follow the usual reviewing process of ESANN,
with an evaluation of originality, technical soundness, and presentation 
style.

In addition, the session organizers also warmly invite authors to 
participate
in an unsupervised NLDR contest. It consists of several data sets with 
various properties
(size, dimensionality, intrinsic dimensionality, manifold or clusters, 
etc.).
Participants are then asked to provide a regular ESANN paper, which 
describes
the method they use, as well as the embeddings for all these data sets.
The quality of each embedding will be assessed by measuring how well
it preserves the K-ary neighborhoods of the data set.
More info and material on https://sites.google.com/site/nldrcontest/

-- 
--

John A. Lee, PhD, FNRS Research Associate
Université catholique de Louvain
Molecular Imaging, Radiotherapy, and Oncology
Avenue Hippocrate 55 box B1.54.07
B-1200 Bruxelles, Belgium
Tel.  +32 2 7649528
Email john.lee at uclouvain.be