Connectionists: CFP: ESANN'2010 special sessions

[esann@dice.ucl.ac.be]

                     ESANN'2010

18th European Symposium on Artificial Neural Networks,
   Computational Intelligence and Machine Learning 

      Bruges (Belgium) - April 28-29-30, 2010

                   Special sessions
======================================================


The following message contains a summary of all special sessions that will
be organized during the ESANN'2010 conference.  Authors are invited to
submit their contributions to one of these sessions or to a regular session,
according to the guidelines found on the web pages of the conference
http://www.dice.ucl.ac.be/esann/.  

Deadline for submissions: November 25, 2009.

According to our policy to limit the number of unsolicited e-mails, we
gathered all special session descriptions in a single message, and try to
avoid sending it to overlapping distribution lists.  We apologize if you
receive multiple copies of this e-mail despite our precautions.  


Special sessions that will be organized during the ESANN'2010 conference
=========================================================

1.  Sparse representation of data
	Thomas Villmann (Univ. Apllied Sciences Mittweida, Germany), 
	Frank-Michael Schleif (Univ. Leipzig, Germany), 
	Barbara Hammer (Clausthal Univ. Of Tech., Germany)

2.  Computational Intelligence in Biomedicine
	Paulo J.G. Lisboa (Liverpool John Moores Univ., U.K.), 
	Alfredo Vellido (Tech. Univ. Catalonia, Spain), 
	José D. Martín (Univ. Valencia, Spain)

3.  Machine learning techniques based on random projections
	Benjamin Schrauwen (Ghent Univ., Belgium), 
	Amaury Lendasse (Helsinki Univ. of Tech., Finland), 
	Yoan Miche (I.N.P. Grenoble, France)

4.  Information Visualization, Nonlinear Dimensionality Reduction, Manifold
and Topological Learning
	Axel Wismüller (Univ. Rochester, New York, USA), 
	Michel Verleysen (Univ. cat. Louvain, Belgium), 
	Michael Aupetit (CEA, France), 
	John Aldo Lee (Univ. cat. Louvain, Belgium)

5.  Computational Intelligence Business Applications
	Thiago Turchetti Maia (Vetta Group, Brazil), Antonio Braga (Univ.
Fed. Minas Gerais, Brazil)

6.  Neuro-Symbolic Reasoning: Theory and Applications
	Massimo De Gregorio (Ist. Cibernetica-CNR, Italy), Priscila M. V.
Lima (Univ. Fed. Rio de Janeiro, Brazil), 
	Gadi Pinkas (Center for Academic Studies, Israel)


Short descriptions
==================

1. Sparse representation of data
-----------------------------------------------------------------------
Organized by:
Thomas Villmann (Univ. Apllied Sciences Mittweida, Germany), Frank-Michael
Schleif (Univ. Leipzig, Germany), Barbara Hammer (Clausthal Univ. Of Tech.,
Germany)

The amount of data available for investigation and analysis is rapidly
growing in various areas of research like in biology/bioinformatics/life
sciences, astronomy/astro physics, physics & chemistry or medicine. Many of
these data sets are very complex such that advanced methods are needed to
extract their inherent but hidden information. Thus, an important task for
data processing is to model these data in an adequate manner while keeping
the models as simple as possible, i.e. a sparse representation of the data
or sparse modelling of the respective underlying problem is demanded.
Thereby, sparseness can be seen in different directions such as
dimensionality/complexity, information optimum modelling, etc. 

Examples for spare concepts are the sparse coding of images introduced by
Olshausen&Field, dimensionality reduction in supervised and unsupervised
learning, relevance learning and feature selection in classification, sparse
prototype representations or sparse wavelet representation and signal
reconstruction, to mention just few. 

Sparseness can be achieved by several methodologies which also may include
additional knowledge about data to be processed, for example the utilization
of functional norms for similarity determination for functional data. 

The session invites to submit paper about topics in adaptive data processing
and machine learning, which explicitly focus on sparseness of data
representation or/and sparse models. 

Example topics could be but are not restricted to 

- the representation of very large data sets, 
- dimensionality reduction, 
- data specific approaches for sparse modelling of special data types like
time series or sequences, spectra, images, 
- information optimum data processing
- sparse prototype models for vector quantization 


2. Computational Intelligence in Biomedicine
-----------------------------------------------------------------------
Organized by: 
Paulo J.G. Lisboa (Liverpool John Moores Univ., U.K.), Alfredo Vellido
(Tech. Univ. Catalonia, Spain), José D. Martín (Univ. Valencia, Spain)

Computational Intelligence techniques (including, broadly, neural networks,
connectionist systems, genetic algorithms, evolutionary programming, fuzzy
systems, and hybrid intelligent systems, according to the scope provided by
the IEEE-CIS) have made significant inroads over the last two decades in the
area of biomedical applications. This is both as beacons of evidence-based
medicine and as robust building blocks of medical decision support systems.
Systems based on Computational Intelligence techniques should have an
important role in defining the methodologies for the next generation of
healthcare delivery technologies. This is expected to follow the 4P
(personalized, predictive, preventive and participatory) agenda, which
demands greater personalization to the needs of the individual patient, a
focus on preventive medicine with the support of predictive approaches, as
well as greater emphasis on pro-active involvement by the patient at the
point of healthcare delivery. This special session aims to be of interest to
CI practioners (with a strong focus on Machine Learning) working in the area
of biomedicine, but also to those biomedicine researchers that have made CI
techniques their tools of choice. The sinergy between both worlds should
guarantee that leading-edge techniques become known to medical practitioners
and that CI research in biomedicine complies with the real-world
requirements of the field. 

The main topics of interest, any of them based on Computational
Intelligence, Machine Learning and otherwise AI-related techniques, include
(but are indeed not necessarily limited to): 

- Methodologies with a focus on model interpretation (including, for
instance, visualization, feature selection and extraction, graphs, and
rules). 
- Structured methodologies for multi-modal data (data fusion combining
different modalities, e.g.: molecular biomarkers, histology, imaging,
electrophysiological measurements and clinical signs). 
- Methodologies for the analysis of functional and spectral signal and
imaging data. 
- Survival analysis. 
- Methodologies for computer-based medical decision support and treatment
planning. 
- Current and planned clinical applications. 
- Methodologies of mining and knowledge discovery applied to medical data. 
- Pharmaceutical research. 

All contributions are meant to strike a reasonable balance between
theoretical novelty and the originality and appropriateness of the
biomedical application.


3. Machine learning techniques based on random projections
-----------------------------------------------------------------------
Organized by:
Benjamin Schrauwen (Ghent Univ., Belgium), Amaury Lendasse (Helsinki Univ.
of Tech., Finland), Yoan Miche (I.N.P. Grenoble, France)

Machine learning techniques based on random projections have recently been
widely used to perform regression, classification and Time Series prediction
tasks. Among the most successful proposed methods lie Reservoir Computing
[1], Extreme Learning Machine [2], Associative Neural Networks [3],
Optimally-Pruned Extreme Learning Machine [4]‚.. 

One of the reasons for the success of random projections based methods is
the extremely good performance in terms of ratio between accuracy and
computational time. Indeed, even though random projection based methods are
often not the most accurate ones, their usual training (learning) time is
orders of magnitude smaller than this of classical methods. Furthermore,
these methods are easily parallelized and can thus benefit from the recent
improvements in the multi-core architecture of modern computers and video
cards. 

Recent developments in random projection led to groundbreaking advances and
different approaches in machine learning. For example, exhaustive and brute
force strategies that were not computationally possible became feasible
within a reasonable time. 

This special session is interested in theoretical advances, new random
projection methods, new learning or meta-learning strategies and industrial
applications for which the ratio accuracy/computational time is crucial. 

REFERENCES: 
[1] David Verstraeten, Benjamin Schrauwen, Michiel D`Haene and Dirk
Stroobandt: An experimental unification of reservoir computing methods
Neural Networks, Vol. 20(3) pp. 391-403 (2007) 
[2] G.-B. Huang, Q.-Y. Zhu and C.-K. Siew: Extreme Learning Machine: Theory
and Applications, Neurocomputing, vol. 70, pp. 489-501, 2006. 
[3] Miller W. T., Glanz F. H., and Kraft L. G. Cmac: An associative neural
network alternative to backpropagation. In Proceedings of the IEEE, volume
70, pages 1561–1567. October 1990. 
[4] Yoan Miche, Patrick Bas, Christian Jutten, Olli Simula and Amaury
Lendasse: A methodology for Building Regression Models using Extreme
Learning Machine: OP-ELM, ESANN 2007, European Symposium on Artificial
Neural Networks, Bruges (Belgium), pages 457-462. April, 2008. 


4. Information Visualization, Nonlinear Dimensionality Reduction, Manifold
and Topological Learning
-----------------------------------------------------------------------
Organized by:
Axel Wismüller (Univ. Rochester, New York, USA), Michel Verleysen (Univ.
cat. Louvain, Belgium), Michael Aupetit (CEA, France), John Aldo Lee (Univ.
cat. Louvain, Belgium)

To extract useful knowledge from exceedingly growing amounts of
high-dimensional data is a ubiquitous challenge throughout science and
engineering. Here, information visualization by topological learning is an
emerging field which is expected to bring new insights in all areas of the
data mining and knowledge discovery process. 

Nonlinear dimensionality reduction methods and manifold learning techniques
are closely related to learning data topology. The differences reside more
in the paradigms (preservation of distances, similarities, or topology) than
in the fundamental goals: visualizing data in 2D or 3D spaces, or embedding
high-dimensional data into lower-dimensional spaces, for information
visualization, data compression, reduction, fighting the curse of
dimensionality, and more. 

Recently there has been significant activity in the field, with the
publication of many methods based for instance on: 

- Inner product preservation by spectral techniques (Isomap, locally linear
embedding, maximum variance unfolding, Laplacian eigenmaps, etc.); 
- Weighted distance or similarity preservation by nonlinear optimization
techniques (Sammon mapping, curvilinear component analysis, stochastic
neighbor embedding, etc.); 
- Self-organization in various types of topology learning networks (Kohonen
feature maps, exploration machines, etc.) Some of these methods can be
applied to non-metric data as well. 

There is no consensus however which method has to be applied in specific
circumstances, or which method performs best in general. More critically,
most of the published works illustrate the methods on toy examples, or on
very few, carefully selected real databases. In addition to these
performance issues, there is no consensus either on which paradigm has to be
followed: distance preservation, topology preservation, self-organization,
other? 
To stimulate the field of nonlinear dimensionality reduction, manifold and
topological learning and its application to information visualization, we
invite paper submissions related, but not limited, to the following
non-exhaustive list of topics: 
- Algorithms, theory and applications of nonlinear dimensionality reduction,
manifold and topological learning 
- Information visualization using manifold and topological learning 
- Spectral clustering and embedding 
- Linear and non-linear dimensionality reduction 
- Links between the above-mentioned paradigms 
- Performance measures for manifold and topological learning 
- Real-world applications of manifold and topological learning throughout
science and engineering, such as in medicine, life and social sciences,
astronomy, economics, and the humanities 


5. Computational Intelligence Business Applications
-----------------------------------------------------------------------
Organized by:
Thiago Turchetti Maia (Vetta Group, Brazil), Antonio Braga (Univ. Fed. Minas
Gerais, Brazil)

There is often a significant research effort from the community to develop
new computational intelligence techniques, but much less effort about how to
effectively deploy these techniques in real-world business applications.
Whenever one goes about applying such techniques to solve existing problems,
one must overcome several practical gaps between theory and effective
applications. In this special session, we welcome contributions with
solutions to these problems. 

Topics of interest include (and are not limited to): 
- Improving the execution time of methods to enhance their applicability in
real problems; 
- Improving the robustness and reliability of methods to handle boundary
conditions in data; 
- Development of new human-machine interfaces meant for non-experts,
allowing the average user to visualize, interact with, and utilize
sophisticated methods; 
- Automatic and semi-automatic model selection; 
- Automatic and semi-automatic parameter tuning; 
- Methodologies and best-practices to collect and prepare data, criteria for
model selection, techniques for parameter tuning, and assessment of model
fitness; 
- Methodologies and best-practices to identify solvable business problems,
evaluate results, and assess returns on investment; 
- New insights in the application of computational intelligence techniques
to real-life business applications; 
- Novel applications of computational intelligence techniques to real-world
problems. 

Authors are encouraged to submit works directly addressing any of the above
(or correlated) items, as well as works on specific business applications
that created innovative solutions to the same problems. Authors are also
encouraged to submit works on novel applications of computational
intelligence to unexplored business problems.


6. Neuro-Symbolic Reasoning: Theory and Applications
-----------------------------------------------------------------------
Organized by:
Massimo De Gregorio (Ist. Cibernetica-CNR, Italy), Priscila M. V. Lima
(Univ. Fed. Rio de Janeiro, Brazil), Gadi Pinkas (Center for Academic
Studies, Israel)

The crucial role of logic in knowledge representation and processing is
unquestionable. However, mechanical reasoning is often a computationally
expensive task, regardless of whether deductive or uncertain reasoning is
effected. Furthermore, most reasoning systems do not exhibit inherent noise
tolerance and learning capabilities. Over the years, sustained efforts have
been made to tackle some of these difficulties by means of Artificial Neural
Networks (ANNs). Still, there is room for improvement in several fronts
ranging from theoretical and methodological issues to real applications, not
to mention efficient solutions to the integration of paradigms.





========================================================
ESANN - European Symposium on Artificial Neural Networks, 
Computational Intelligence and Machine Learning 
http://www.dice.ucl.ac.be/esann

* For submissions of papers, reviews, registrations:
Michel Verleysen
Univ. Cath. de Louvain - Machine Learning Group 
3, pl. du Levant - B-1348 Louvain-la-Neuve - Belgium
tel: +32 10 47 25 51 - fax: + 32 10 47 25 98 
mailto:esann at uclouvain.be

* Conference secretariat
d-side conference services
24 av. L. Mommaerts - B-1140 Evere - Belgium
tel: + 32 2 730 06 11 - fax: + 32 2 730 06 00 
mailto:esann at uclouvain.be
========================================================