Connectionists: ICHSA workshop: new paradigms for hybrid learning systems

[Dario Malchiodi]

          Sorry, if you received this message more than once.
     Feel free to distribute it to those who might be interested.

################################################################

                                            CALL FOR CONTRIBUTIONS

                                                       WORKSHOP ON

                       NEW PARADIGMS FOR HYBRID LEARNING SYSTEMS

http://laren.dsi.unimi.it/new_paradigm


within  the International Conference on Hybrid Systems and Applications
(ICHSA 2006)
May 24th 2006, The University of Louisiana, Lafayette, LA, USA

http://cos.fit.edu/math/ichsa/



Submission deadline: February 15th 2006

################################################################

Organizers:
------------

  * Bruno Apolloni, University of Milano
    http://laren.dsi.unimi.it/apolloni
  * Zong Sha, Chinese Institute of electronics,Beijing, China
    zsha at public.bat.net.cn, z.sha at ieee.org
  * Dario Malchiodi, University of Milano
    http://homes.dsi.unimi.it/~malchiod



Besides the common dichotomy between subsymbolic learning systems such 
as neural networks and symbolic systems such as decision trees, and 
their various forms of hybridization, new algorithms are raising for 
learning, sharing features of both symbolic and subsymbolic paradigms. 
Of classification algorithms like SVM or feature extraction algorithms 
like ICA you cannot say to be completely symbolic since the kernels/non 
gaussianity measures selection must be drawn by data in a non direct 
way -- guessed in any way -- while the goal they optimize is an 
explicit function of the parameters they aim to learn.
In very broad terms the object of this special session is to gather 
various approaches to learning, where the distinction between what 
comes from axiomatic theories and what is left to the ability of the 
learner and his heuristics is untenable. Rather we may distinguish 
between different strategies with which the users organize past data in 
order to face their continuation.


Contributions:
--------------
As it is, the scope of this special session is open to the contribution 
of researchers in many fields, ranging from statistics to granular 
computing, neural networks, evolutionary computation, computational 
learning and so on.  A definite preference is for non conventional 
approaches, provided a clear rationale and either formal proofs or 
stringent numerical results are supplied.

We expect extended abstracts six to eight pages long without special 
format. Authors of the accepted contributions will be invited to submit 
a paper in the format requested by the conference proceedings’ editor.

Please submit extended abstracts to new_paradigms at dsi.unimi.it within 
February 15th 2006.

Important dates:
-------------------

February 15th 2006 Submission deadline
March 15th 2006 Notification of acceptance
April 15th 2006 Camera ready papers deadline
May 22nd 2006 Conference