PRESS RELEASE: Abbadingo One Machine Learning Competition - ANNOUCEMENT
CONTACT: Barak Pearlmutter, 505 277-9425, abbadingo@cs.unm.edu
RELEASE DATE: MARCH 7, 1997 (REVISED: MARCH 18, 1997)
Abbadingo One: DFA Learning Competition
This machine learning challenge problem is aimed at anyone interested
in making computers extract underlying structure from masses of data.
Current machine learning techniques are inadequate to the increasing
volumes of information whose structure we would like to induce, and to
the increasing complexity of the underlying regularities. In order to
encourage the development of algorithms that can handle larger
problems, and can correctly induce more complicated things from less
data, we have organized a competition.
A table of twelve problems, each beyond the current state of the art,
are being made publicly available. Whoever solves a previously
unsolved problem harder than any solved problem (before the end of the
competition on 15-Nov-1997) will receive an award of at least $1,024.
We have stripped the problem down to its essentials: find the
structure of a regular language (a very restricted but common class of
formal languages) based on a set of training samples classified by a
randomly generated deterministic finite automaton --- and demonstrate
competence by classifying a set of test strings with over 99%
accuracy. All of the twelve problems are identical in form, differing
only in the amount of data available and the size of the target DFA.
The competition is being sponsored by, among others,
* The Computer Science Department at the University of New Mexico,
which is providing computational support.
* The prestigious Kluwer Academic journal "Machine Learning," which
will give priority treatment to a paper describing the award
winning algorithm.
* The Santa Fe Institute, which will host the award ceremony.
* The "Journal of Artificial Intelligence Research."
For details retrieve http://abbadingo.cs.unm.edu/
May the best algorithms win!
--
Competition organizers:
Kevin J. Lang, Research Scientist, NEC Research Institute, Princeton NJ
Barak A. Pearlmutter, Assistant Prof, Comp Sci Dept, Univ New Mexico