ESANN'2005
13th European Symposium
on Artificial Neural Networks
Bruges (Belgium) - April 27-28-29, 2005
Announcement and call for papers
=====================================================
The following message contains a summary of all special sessions that
will be organized during the ESANN'2005 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/).
Special sessions that will be organized during the ESANN'2005 conference
========================================================================
1. Kernel methods and the exponential family
A. Smola, National ICT Australia, ANU (Australia) &
S. Canu, PSI, CNRS - INSA Rouen (France)
2. Classification using non-standard metrics
B. Hammer, Univ. Osnabrück, T. Villmann, Univ. Leipzig (Germany)
3. Evolutionary and neural computation
C. Igel, Ruhr-Univ. Bochum, B. Sendhoff, Honda Research Institute
Europe GmbH (Germany)
4. Dynamical and Numerical Aspects of Neural Computing
M. Atencia, University of Málaga (Spain)
5. Artificial Neural Networks and Prognosis in Medicine
José M. Jerez, Univ. Malaga (Spain), Leonardo Franco, Univ. Oxford
(UK)
Short descriptions
==================
Kernel methods and the exponential family
-----------------------------------------
Organized by :
- A. Smola, National ICT Australia, ANU (Australia)
- S. Canu, PSI, CNRS - INSA Rouen (France)
The success of Support Vector Machine (SVM) gave rise to the development
of a new class of theoretically elegant learning machines which use a
central concept of /kernels/. Exponential families, a standard tool in
statistics, can be used to unify many existing machine learning
algorithms based on /kernels/ (such as SVM) and to invent novel ones
quite effortlessly.
This session aims at bringing together researchers interested in all
aspects of kernel machines to meet and share their new ideas in
particular on the relationship between kernels methods for machine
learning and exponential families.
Interesting topics includes the following non-exhaustive list: Support
Vector Machines, Gaussian Processes, Conditional Random Fields,
definition of similarity measures, optimization, Kernel PCA, sparse
estimation procedures, kernel clustering, hyperkernels, kernels on
graphs and dynamical systems, string kernels. This includes also
applications for instance in the fields of pattern recognition, text
analysis, computer vision, context aware programming, signal processing,
novelty detection and network security.
Classification using non-standard metrics
-----------------------------------------
Organized by:
- B. Hammer, Univ. Osnabrück (Germany)
- T. Villmann, Univ. Leipzig (Germany)
Supervised and unsupervised classification problems play a major role in
machine learning and many different neural classification algorithms
such as SOM or LVQ exist. Most algorithms, however, heavily rely on the
Euclidean metric and they are therefore best suited for real vectors of
a fixed and finite dimensionality. Data from interesting application
areas such as language, spectrometry or bioinformatics are very high
dimensional, noisy, and heterogeneous, or they possess additional
correlations and further structural elements such as time dependencies.
Algorithms based on the Euclidean metric often fail in these scenarios.
This session will focus on classification methods which use specifically
designed metrics for these type data such as learning metrics,
specifically designed kernels, or discrete metrics. Submissions are
encouraged within (but not restricted to) following areas:
- metric adaptation and learning metrics
- clustering using discrete metrics
- specifically designed kernels and metrics
- classification of very high dimensional data
- classification of temproal or structured data
- applications e.g. in spectrometry or bioinformatics
Evolutionary and neural computation
-----------------------------------
Organized by:
- C. Igel, Ruhr-Univ. Bochum
- B. Sendhoff, Honda Research Institute Europe GmbH (Germany)
Evolutionary computation has proven to be a competitive approach to the
adaptation of neural networks, in particular in scenarios where
"classical" optimization methods are not applicable. The most prominent
examples are model selection (e.g, for multi-layer perceptrons, radial
basis function networks, and support vector machines), the field of
evolutionary robotics, computational neuroscience, and reinforcement
learning (in particular evolving strategies for games). We welcome
papers describing theoretical work and applications in these and related
fields. Papers dealing with applications and empiricial analysis should
include statistical tests and (where possible) a comparison with
alternative state of the art methods.
Dynamical and Numerical Aspects of Neural Computing
---------------------------------------------------
Organized by:
- M. Atencia, University of Málaga (Spain)
Over the last decade, the synergy between dynamical systems theory and
numerical analysis has contributed to the development of several
branches of mathematics, such as geometric integration, optimization and
control. Many of these techniques are still far from being widely spread
within the computer science community. This session will aim at gaining
further insight in the analysis of neural networks by incorporating
concepts from systems theory and novel numerical methods. On one hand,
recurrent networks are dynamical systems that can be analysed with the
techniques of systems theory. On the other hand, the implementation of
neural networks often requires numerical methods that preserve the
qualitative properties of continuous models. Particularly welcome are
submissions that emphasize the mutual benefits of a multidisciplinary
approach, within the following non-exhaustive list of topics:
- Analysis of recurrent networks: stability, bifurcations and chaos. -
Computational complexity.
- Infinite dimensional systems: stochastic, systems with delays, models
with partial derivatives. Differential equations on manifolds.
- Novel numerical methods of wide applicability. Connection between
neural networks and conventional methods, such as homotopy methods,
interior point algorithms, etc.
- Stability and convergence under discretization. Implementation issues,
error control, convergence speed.
- Applications: control engineering, optimization, etc.
Artificial Neural Networks and Prognosis in Medicine
----------------------------------------------------
Organized by:
- José M. Jerez, Univ. Malaga (Spain)
- Leonardo Franco, Univ. Oxford (UK)
Prognosis can be defined as an estimate of cure, recurrence of disease,
or survival for a patient. Frequently, physicians are asked for
prognostic assessments and often worry that their assessments will prove
inaccurate. In this situation, the analysis of the prognostic
information appears as an important part of the medical care, since its
influences decisions regarding individual preferences, treatments, and
resources allocation. This situation has led the scientific community to
develop prognostic system based on risk factors, staging systems,
decision rules, statistical models and computer algorithms. During the
last years, the application of artificial neural networks for prognostic
of a patient outcome in clinical medicine has attracted growing interest
in medical and bioinformatic research areas. The session aims to foster
exchange of ideas among researchers involved in medical prognosis, with
a special emphasis to promote discussions about the interaction and
comparison between neural networks, genetic algorithm, decision trees,
statistical methods, etc.
Other topics of interest (related to prognosis in medicine) include:
- Treatment of missing data
- Data preprocessing and data extraction methods
- Generation of interpretable rules from non-symbolic representations
- Probabilistic Models, Bayesian Networks, Hidden Markov Model
- Appropriate statistical tools to evaluate prognosis accuracy
========================================================
ESANN - European Symposium on Artificial Neural Networks
http://www.dice.ucl.ac.be/esann
* For submissions of papers, reviews,...
Michel Verleysen
Univ. Cath. de Louvain - Microelectronics Laboratory
3, pl. du Levant - B-1348 Louvain-la-Neuve - Belgium
tel: +32 10 47 25 51 - fax: + 32 10 47 25 98
mailto:esann@dice.ucl.ac.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@dice.ucl.ac.be
========================================================
|
Home Login Meetings Courses Belgium Maillist Credits Servers Links Archive
Please send comments and suggestions to
Mike Wijnants. |
|
Page last updated on © BBF 1998 all rights reserved |