ESANN'2009
17th European Symposium on Artificial Neural Networks
Advances in Computational Intelligence and Learning
Bruges (Belgium) - April 22-23-24, 2009
Special sessions
=============================================
The following message contains a summary of all special sessions that will
be organized during the ESANN'2009 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 21,
2008.
Special sessions that will be organized during the ESANN'2009 conference
=========================================================
1. Semi-supervised learning
Antônio de Pádua Braga (Federal Univ. Minas Gerais, Brazil)
2. Learning (with) Preferences
Fabio Aiolli, Alessandro Sperduti (Univ. degli Studi di Padova, Italy)
3. Brain Computer Interfaces: from theory to practice
Luc Boullart (Ghent University), Patrick Santens (Ghent University Hospital),
George Otte (Dr. Guislain Institute), Bart Wyns (Ghent University, Belgium)
4. Efficient learning in recurrent networks
Benjamin Schrauwen (Ghent University, Belgium), Jochen J. Steil (Bielefeld
University, Germany), Barbara Hammer (Clausthal University of Technology,
Germany)
5. Weightless Neural Systems
Massimo De Gregorio (Istituto di Cibernetica-CNR, Italy), Priscila M. V. Lima,
Felipe M. G. França (Universidade Federal do Rio de Janeiro, Brazil)
6. Neural Maps and Learning Vector Quantization - Theory and Applications
Thomas Villmann, Frank-Michael Schleif (Univ. Leipzig, Germany)
Short descriptions
==================
1. Semi-supervised learning
-----------------------------------------------------------------------
Organized by:
Antônio de Pádua Braga (Federal Univ. Minas Gerais, Brazil)
Semi-Supervised learning falls in-between the Supervised and Unsupervised
Learning paradigms, by considering both labeled and unlabeled data for
training. From the Supervised Learning perspective, structural information,
particularly related to the separation margin, is usually added to the
Optimization problem resulted from the labeled data. From the Unsupervised
Learning perspective, Semi-Supervised Clustering is accomplished by
considering the labeled data as constraints to the clustering task. Despite
of having different goals, the basic elements of training in both
perspectives are the labels plus structural information obtained from the
unlabeled data set. In this special session, we seek for contributions from
both perspectives above, not limited to Artificial Neural Networks design.
Topics of interest include (but are not limited to):
- Semi-supervised learning
- Semi-supervised clustering
- Transductive learning
- Co-training
- Partial supervision
2. Learning (with) Preferences
-----------------------------------------------------------------------
Organized by:
Fabio Aiolli, Alessandro Sperduti (Univ. degli Studi di Padova, Italy)
Preferences give a declarative way for specifying desires and are very
important in many applications which include reccomender systems for
e-commerce and social networks, ranking systems for information retrieval,
and player modelling for games. In all these contexts, people find easier to
indicate which objects they prefer to which other with respect to make
absolute judgments about the relevance they give to each of them.
Recently, preference learning models and preference based predictions have
gained popularity in the machine learning and knowledge discovery
communities. Many supervised learning tasks can in fact be modeled as sets
of preferences over a parameterized relevance function. This kind of
preferences are given in the form of partial or full orders over the
relevance function. Preferences can be given between objects (instance
rankings) and/or between classes (label rankings).
Other interesting topics concern how to mine or elicitate preferences from
user behaviours and how to aggregate preferences obtained from multiple
sources.
We invite papers on learning preferences and/or learning with preferences.
In particular topics of interest include, but are not limited to:
- theory about any aspect of preference learning
- preference based models to cope with structured (complex) predictions
- preference mining and preference elicitation
- preference/ranking aggregation
- semi-supervised preference learning
- scalability and efficiency of preference based learning algorithms
- evaluation measures for preference learning
- applications of preference learning: information retrieval, e-commerce, games, ecc.
Submitted papers will be reviewed according to the ESANN reviewing process
and will be evaluated on their scientific significance, originality,
correctness, and writing style.
3. Brain Computer Interfaces: from theory to practice
-----------------------------------------------------------------------
Organized by:
Luc Boullart (Ghent University), Patrick Santens (Ghent University
Hospital), George Otte (Dr. Guislain Institute), Bart Wyns (Ghent
University, Belgium)
Brain-Computer Interfaces (BCI) are a new kind of human-machine interfaces
and represents a burgeoning field of research. Brain signals are measured
using EEG and translated directly into control commands. A typical
application of BCI is found in people with severe motor disabilities
allowing them to manipulate their environment in an alternative way. However
there?s still a lot of work to be done to make it usable in daily life.
This special session aims at presenting novel ideas of brain signal
analysis, artefact removal algorithms (for example blind source separation),
feature selection strategies and BCI classification algorithms or
interesting applications of BCI for robot control.
Keywords: (brain) signal processing and modelling, brain-computer
interfaces, intelligent ?brain? controlled computers, EEG signal analysis
4. Efficient learning in recurrent networks
-----------------------------------------------------------------------
Organized by:
Benjamin Schrauwen (Ghent University, Belgium), Jochen J. Steil (Bielefeld
University, Germany), Barbara Hammer (Clausthal University of Technology,
Germany)
Recurrent neural networks carry the promise of efficient biologically
plausible signal processing models optimally suited for a wide area of
applications, especially when dealing with spatio-temporal data or
causalities. On the other hand, they can form the basis for an explanation
for neurophysiological processes and cognitive phenomena of the human brain.
Recently, a number of fundamental paradigms connected to RNNs have been
developed which allow new insights into potential supervised and
unsupervised information processing with RNNs and open the way to new
efficient training algorithms which overcome the well-known problems of
long-term dependencies. The aim of the session is to further the
understanding and development of efficient, biologically plausible recurrent
information processing, both in theory and in applications.
Submissions are encouraged within the following non-exhaustive list of
keywords:
- reservoir computing: echo state machine, liquid state machine
- recurrent SOM
- LSTM
- unsupervised and semi-supervised adaptation of RNNs
- evolutionary models for RNNs
- connection of RNNs and brain phenomena
- connection of RNNs and symbolic reasoning
- theory of RNN dynamics, learning, and generalization
- applications
5. Weightless Neural Systems
-----------------------------------------------------------------------
Organized by:
Massimo De Gregorio (Istituto di Cibernetica-CNR, Italy), Priscila M. V.
Lima, Felipe M. G. França (Universidade Federal do Rio de Janeiro, Brazil)
Mimicking biological neurons by focusing on the decoding performed by the
dendritic trees is a different and attractive alternative to the
integrate-and-fire McCullogh-Pitts neuron stylisation. RAM-based or Boolean
neurons and systems have been studied and applied in a wide spectrum of
situations.
This session invites original contributions on theoretical and practical
aspects of weightless neural systems, at all levels of abstraction (pattern
recognition, consciousness, artificial emotions, reasoning etc).
6. Neural Maps and Learning Vector Quantization - Theory and Applications
-----------------------------------------------------------------------
Organized by:
Thomas Villmann, Frank-Michael Schleif (Univ. Leipzig, Germany)
Neural maps and learning vector quantization constitute important neural
paradigms in unsupervised and supervised vector quantization. Prominent
methods are the self-organizing map (SOM), neural gas (NG) and the family of
LVQ-algorithms or generalizations thereof. Although most of the approaches
are well-known, there are still open theroretical questions like
magnification for Heskes-SOM or non-euclidean NG, the dynamics of LVQ, to
name just a few. Recent investigations and extensions are in the field of
non-standard metrics, structured data processing, time series, batch and
patch-variants etc. All these interesting new developments lead to a broader
range of applications of the algorithms compared to their standard variants.
The proposed session invite researchers to submit contribution about new
approaches, extensions and modifications as well as ideas in this outlined
direction. Thereby, new theoretical investigations as well as outstanding
applications demonstrating the abilities of new extensions/modifications of
the standard algorithm are in the focus. For the latter aspect a strong
connection between the specific aspects of SOM/NG/LVQ to the application
should be explicitely given and highlighted.
Submissions are encouraged within the following non-exhaustive list of
topics:
- theory of SOM/NG/LVQ and variants thereof
- magnification and magnification control
- non-standard metrics
- new extensions of existing approaches
- semi-supervised learning
- fuzzy methods for neural maps
- statistical interpretations
- learning theory
- outstanding applications
========================================================
ESANN - European Symposium on Artificial Neural Networks -
Advances in Computational Intelligence and 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@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@uclouvain.be
========================================================
Show Posts
Recent
In 
