Connectionists: Workshop announcements

[John Shawe-Taylor]

Please note that two PASCAL workshops are planned back to back in early
July:

 


Statistics and <http://www.pascal-network.org/Workshops/SOC05/>
Optimisation of Clustering Workshop


4-5th July, 2005, Cumberland Lodge, Windsor, UK


Organisers: John Shawe-Taylor
<http://www.pascal-network.org/Network/Researchers/1/> 

University <http://www.pascal-network.org/Network/Sites/1/>  of Southampton


Shai Ben-David <http://www.pascal-network.org/Network/Researchers/496/> 

University of Waterloo, Canada

Motivated by growing demand from many practical data management
applications, clustering has recently emerged as a central research
direction. While this research has resulted in many novel and sophisticated
algorithms, the theory to support and evaluate these algorithms has been
building up at a much slower pace. Common clustering techniques make a
fundamental reliance on heuristics both in the definition of the task and
the algorithms used to solve it.

Many of the clustering tasks that have been formally analyzed turn out to be
NP hard, and most of the theoretical work in this area has focused on
finding polynomial time approximation algorithms for hard clustering
problems. 

While such algorithms may work well for moderately sized data sets, the
pattern of exponential growth of data sets suggests that even linear time
algorithms may soon be too slow in practice. The next step is to consider
sample-based clustering algorithms that rely on a statistical analysis of
the problem. Such algorithms can be sub-linear since they can output a
clustering of the data without being able to even read the entire data set. 

This workshop will bring together researchers from Machine Learning
Statistics and Optimization Theory that have recently been developing
algorithmic and statistical foundations for clustering. We wish to provide
an opportunity for mutual exposure and interaction to a variety of research
approaches that share the goal of providing the high efficiency required by
current clustering applications while retaining a sound mathematical
justification. 

In accordance with the PASCAL workshop programme, the workshop will be
recorded on video and all presentations will be made available to PASCAL
researchers. 

It is anticipated that contributions to the workshop will be published in an
edited collection of papers. In conjunction with an agreement between the
network and Springer-Verlag, the organisers are currently organising for the
inclusion of the proceedings of the workshop in the Lecture Notes in
Computer Science series. 

The workshop will include some invited contributions listed below, but also
solicits submissions addressing the topics covered in the workshop. 

Invited contributions: Shai Ben-David, Olivier Bousquet, Joachim Buhmann,
Moses Charikar, 

For more details follow the links.  

 


Principled methods of <http://www.pascal-network.org/Workshops/TEE05/>
trading exploration and exploitation Workshop


6-7th July, 2005, Cumberland Lodge, Windsor, UK


Organisers:

 


Peter Auer <http://www.pascal-network.org/Network/Researchers/276/> 

University <http://www.pascal-network.org/Network/Sites/47/>  of Leoben


Nicolò Cesa-Bianchi <http://www.pascal-network.org/Network/Researchers/53/> 

Università <http://www.pascal-network.org/Network/Sites/8/>  degli Studi di
Milano


Colin de la <http://www.pascal-network.org/Network/Researchers/219/>
Higuera

Université <http://www.pascal-network.org/Network/Sites/38/>  Jean Monnet -
Saint-Etienne


John Shawe-Taylor <http://www.pascal-network.org/Network/Researchers/1/> 

University <http://www.pascal-network.org/Network/Sites/1/>  of Southampton


Chris Watkins <http://www.pascal-network.org/Network/Researchers/298/> 

Royal <http://www.pascal-network.org/Network/Sites/51/>  Holloway,
University of London

Traditional off-line learning methods are often not appropriate for
applications in user modelling and user interfaces since to be useful the
system must learn about the user or context during the process of
interaction 'on the fly'. This immediately raises the fundamental problem of
trading off exploration and exploitation in that as information is learnt
the system may be tempted to act in line with this insight rather than
further exploring alternatives. 

Machine learning has developed a number of models that attempt to capture
and analyse this trade-off, from the simplest bandit problem to the full
Markov decision processes underlying reinforcement learning. 

The workshop would include tutorials covering the bandit analysis as well as
its relevance to user modelling. Reinforcement learning would also be
included with particular emphasis on applications in user interfaces. It is
also hoped to launch a challenge in this area. 

This workshop comes under the Thematic Programme 4: On-line User Modelling
and Reinforcement Learning and is a core meeting of the PASCAL Network. 

In accordance with the PASCAL workshop programme, the workshop will be
recorded on video and all presentations will be made available to PASCAL
researchers. 

It is anticipated that contributions to the workshop will be published in an
edited collection of papers. In conjunction with an agreement between the
network and Springer-Verlag, the organisers are currently organising for the
inclusion of the proceedings of the workshop in the Lecture Notes in
Computer Science series. 

Invited contributions: Nicolo Cesa-Bianchi, Peter Auer, and further speakers
TBA. 


 


For more information about the PASCAL Network please follow the link:
<http://www.pascal-network.org/> http://www.pascal-network.org .