special sessions at ESANN'2004
Michel Verleysen
verleysen at dice.ucl.ac.be
Wed Oct 29 11:33:03 EST 2003
----------------------------------------------------
| |
| ESANN'2004 |
| |
| 12h European Symposium |
| on Artificial Neural Networks |
| |
| Bruges (Belgium) - April 28-29-30, 2004 |
| |
| Special sessions |
----------------------------------------------------
The following message contains a summary of all special sessions that will
be organized during the ESANN'2004 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/).
List of special sessions that will be organized during the ESANN'2004
conference
=====================================================================
1. Neural methods for non-standard data
B. Hammer, Univ. Osnabr=FCck, B.J. Jain, Tech. Univ. Berlin (Germany)
2. Soft-computing techniques for time series forecasting
I. Rojas, H. Pomares, Univ. Granada (Spain)
3. Neural networks for data mining
R. Andonie, Central Washington Univ. (USA)
4. Theory and applications of neural maps
U. Seiffert, IPK Gatersleben, T. Villmann, Univ. Leipzig,
A. Wism=FCller, Univ. Munich (Germany)
5. Industrial applications of neural networks
L.M. Reyneri, Politecnico. di Torino (Italy)
6. Hardware systems for Neural devices
P. Fleury, A. Bofill-i-Petit, Univ. Edinburgh (Scotland, UK)
Short descriptions
==================
Neural methods for non-standard data
------------------------------------
Organized by :
- B. Hammer, Univ. Osnabr=FCck (Germany)
- B.J. Jain, Tech. Univ. Berlin (Germany)
In modern neural network research it is common practice to represent data as
feature vectors in an Euclidean vector space. This kind of representation is
convenient; due to possibly high dimensions or potential loss of structural
information, however, it is limited for many relevant application areas
including bioinformatics, chemistry, natural language processing, network
analysis, or text mining. Alternative powerful and expressive
representations for complex data structures include, for example, graphs,
trees, strings, sequences, or functions. Recent neural models which directly
deal with complex data structures include recursive models, kernels for
structures, or functional networks, to name just a few approaches. The
session will focus on neural techniques for processing of non-vectorial
data. Authors are invited to submit contributions related to the following
list of topics:
- supervised and unsupervised models for complex data structures,
- coupling of symbolic and sub-symbolic systems,
- similarity measures and kernel models for non-vectorial data,
- specific preprocessing methods for complex data structures,
- incorporatation of prior knowledge and invariances,
- theoretical results within this topic,
- applications e.g. in bioinformatics, chemistry, language processing,
- time series processing, graph processing.
Soft-computing techniques for time series forecasting
-----------------------------------------------------
Organized by :
- I. Rojas, Univ. Granada (Spain)
- H. Pomares, Univ. Granada (Spain)
It is obvious that forecasting activities play an important role in our
daily life. A time series is a sequence of measured quantities, of some
physical system taken at regular intervals of time. Time series analysis
includes three important specific problems: prediction, modelling, and
characterization. The goal of prediction is to accurately forecast the
short-term evolution of the system, the aim of modelling is to precisely
capture the features of the long-term behaviour of the system, and the
purpose of system characterization is to determine some underlying
fundamental properties of the system. Papers concerning these goals, using
traditional statistical model (ARMA), neural networks, soft-computing
techniques, fuzzy system, evolutionary algorithms, etc are welcome.
Neural networks for data mining
-------------------------------
Organized by :
- R. Andonie, Central Washington Univ. (USA)
Data mining is an attractive application area for neural networks. This
session will focus on the specificity and limits of neural computation for
data mining. The following questions will be discussed:
1. What makes the difference between data for data mining applications and
data for other NN applications: huge data bases, mixed data types, uncertain
data, redundant and conflicting data, etc.
2. Data mining applications may be related to Internet applications with
on-line processing capability. Only few NN models can handle such
requirements. How useful are NN in this case?
3. Data mining includes not only knowledge acquisition (rule extraction) but
also decision making. This can be done by using NN models. How specific is
this task, considering points 1-2?
4. Computational intelligence applications in E-commerce, customer
profiling, marketing segmentation, etc.
Authors are invited to submit contributions related to neural and
neuro-fuzzy techniques used in data mining. Papers discussing why/when/how
neural models are appropriate for data mining applications are especially
welcome.
Theory and applications of neural maps
--------------------------------------
Organized by :
- U. Seiffert, IPK Gatersleben (Germany)
- T. Villmann, Univ. Leipzig (Germany)
- A. Wism=FCller, Univ. Munich (Germany)
Neural maps in real biological systems can be seen as information processing
systems which map complex information onto a roughly two-dimensional
structure such that the statistical relations within the data are
transformed into geometrical relations - called topographic mapping. Models
which describe these brain properties are called neural maps. In technical
context these models are utilized as topographic vector quantizers. Famous
examples are the Self-Organizing Map (SOM), the Elastic Net (EN), etc.
However, also other vector quantizers, originally not inspired by biological
motivation, can be taken as topographic vector quantizers. Examples are the
Neural Gas (NG), Soft Topographic Vector Quantization (STVQ) and other.
Topographic vector quantizers have found a large range of applications in
data mining, visualization, data processing, control and so on. In parallel,
a growing number of extensions of existing algorithms as well as new
approaches were developed during the last years. In the proposed session we
want to focus onto new developments of topographic vector quantization and
neural maps. Thereby we emphasize the theoretical background as well as
interesting applications with key ideas for optimal use of the properties of
neural maps. We invite researchers to provide new ideas in these topics.
Possible contributions can be in any area matching this framework with
following (but not restricted) topics:
- theory of topographic vector quantization
- estimation of probability density
- image processing
- time series prediction
- classification tasks
- pattern classification, clustering, fuzzy clustering
- blind source separation and decorrelation
- dimension and noise reduction
- evaluation of non-metric data (categorical/ordinal)
- data mining
Industrial applications of neural networks
------------------------------------------
Organized by :
- L.M. Reyneri, Politecnico. di Torino (Italy)
Hardware systems for Neural devices
-----------------------------------
Organized by :
- P. Fleury, Univ. Edinburgh (Scotland, UK)
- A. Bofill-i-Petit, Univ. Edinburgh (Scotland, UK)
This special session aims to present new developments in neural hardware
engineering to the neural networks community. The emphasis will be placed on
the computational properties of the hardware systems or their use in
specific applications, rather than on intricate details of circuit
implementations. Some suggested areas of interest include (but are not
restricted to) neuromorphic VLSI, interfaces between biological neurons and
hardware, implementation of novel ANN algorithms and stochastic computing
with nanotechnologies.
========================================================
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 at 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 at dice.ucl.ac.be
========================================================
More information about the Connectionists
mailing list