Probabilistic Automata learning Competition

PAutomaC is a competition about learning probabilistic finite state models (PFA, HMM, WA, …), where participants are to build algorithms that can learn rational distribution over strings.

On its web page (http://snowball.cs.umbc.edu/icgi2012/challenge/), PAutomaC will allow users to download data samples corresponding to several kind of targets : some are deterministic, some not, the size of the alphabet is varying, and so is the complexity of the model.

There will be two kind of data: artificial ones that are obtained from a randomly generated machine; real data from different fields.

Schedule:

– End of January: PAutomaC is open, anyone may register, develop and test their algorithms on training data sets
– May 20th: Competition data sets are available
– June 30th: Competition is over
– July 20th: Short papers are submitted
– September 12-15th: special session at ICGI’12 in Washington, DC, USA (http://www.coral-lab.org/icgi2012/ )

Prizes and publications:

We are quite confident to find a sponsor that will award a prize to the winner of PAutomaC.
Participants are encouraged to submit an extended abstract and to present their innovations at the PAutomaC special session that will be organised during ICGI.

A submission to a journal special issue is also being considered.

Scientific committee:

– Peter Adriaans, University of Amsterdam, The Netherlands

– Dana Angluin, Yale University, USA

– Pierre Dupont, Université catholique de Louvain, Belgium.

– Ricard Gavalda, Universitat Politècnica de Catalunya, Spain

– Colin de la Higuera, University of Nantes, France

– Jean-Christophe Janodet, University of Evry, France

– Tim Oates, University of Maryland, USA

– Jose Oncina, University of Alicante, Spain

– Menno van Zaanen, Tilburg University, The Netherlands

A further call for participation will detail the running and the technical aspects of the competition.