Connectionists: special session on reservoir computing at IJCNN

[Peter Tino]

2010 International Joint Conference on Neural Networks (IJCNN'10)
Call for Papers: Special Session on Reservoir Computation

Scope:
Recently there has been an outburst of research activity in the field of
reservoir computation. Reservoir based techniques rely on dynamical models
for processing time series that make a conceptual separation of the
temporal data
processing into two parts: 1) representation of temporal structure in the
input stream through a non-adaptable dynamic "reservoir", and 2) a
memoryless easy-to-adapt readout from the reservoir. Such models have been
successfully used in many application domains, for example time-series
prediction, speech recognition, noise modelling, dynamic pattern
classification, reinforcement learning and language modelling.
The field of reservoir computation has been growing rapidly with dedicated
special sessions at conferences and special issues of journals. However,
reservoir computation has been rightfully criticized for not being
principled enough. There have been several attempts to address the question
of what exactly is a `good' reservoir for a given application, but no
coherent theory has yet emerged. The largely black-box character of
reservoirs prevents a deeper theoretical investigation of the dynamical
properties of successful reservoirs. Reservoir construction is largely
driven by a series of
(more-or-less) ad-hoc randomized model building stages, with both the
researchers and practitioners having to rely a series of trials and errors.
Often reservoirs have been evolved in a costly and difficult-to-analyze
evolutionary computation setting.
The aim of the special session is to conceptualize the field of reservoir
computation in the context of recent developments, so as to enable its
systematic study.

Topics
In particular we encourage submissions addressing the following issues:

     ·    Measures of the `reservoir quality' .
     ·    Putting reservoir models in the context of other state-of-art
learning methods operating on temporal data .
     ·    Applications of reservoir computation .
     ·    New extensions/formulations of reservoir models .
     ·    Combination of reservoir models with other statistical or
natural computation techniques .
     ·    Theoretical analysis of reservoirs.


Paper submission
All instructions and templates for submission can be found in the WCCI'10
web site (http://www.wcci2010.org/submission). Please, contact to the
special session organizers if you are planning to submit any paper.


Important Dates
Deadline for paper submission                  January 31, 2010
Notification of acceptance                     March 15, 2010
Final Paper Submission                         May 2, 2010


Organizers
- Peter Tio, School of Computer Science, University of Birmingham, UK.
  P.Tino at cs.bham.ac.uk

- Pedro A. Gutierrez, Dept. of Computer Science and Numerical Analysis,
University of Cordoba, Spain.
  pagutierrez at uco.es

- Ali Rodan. School of Computer Science, University of Birmingham, UK
 A.A.Rodan at cs.bham.ac.uk


-- 
Peter Tino 
The University of Birmingham
School of Computer Science 
Edgbaston, Birmingham B15 2TT, UK 
+44 121 414 8558
http://www.cs.bham.ac.uk/~pxt/