June 25th: the second real data problem is available. It corresponds to size 20 sliding windows over a discretized sensor signal.
June 5th: the first real data problem is available. It corresponds to POS sequences coming from sentences of a large corpus.
May 29th: the last 12 artificial problems are available.
May 20th: Phase 2 is launched: The data of the real competition are available! The data sets of the training phase are still available but you cannot submit your results any more. However, the files containing the true probabilities (obtained with the target automata) are available.
May 2nd: The competition is now officially sponsored by the european network of excellence PASCAL2!
May 1st: 30 new problems are available in the download section. They correspond to bigger and harder problems.
April 16th: implementations of the ALERGIA and of the Baum-Welch algorithms are available in the Baselines section.
April 2nd: 4 new data data sets are available. They correspond to PDFA targets while the previous ones (problems numbered from 10 to 13) were actually big PNFA...
April 1st: a facebook page for the competition is available. Participants can discuss there about the competition.
March 30th: a Hall of fame is available. The names of participants whose submissions are in the the top 3 is given for each problem.
March 26th: an article describing the competition and its context is now available.
March 20th: 16 new data sets are available!
March 10th: CASL is officially sponsoring the competition. They will offer a $500 prize to the winner!
March 8th: The website is fully operational, the first data set is available
|February 2012||Base learner(s) and first testing data-sets are available.|
|March 2012||Full testing data-sets are provided and the submission site is open for testing your algorithm(s).|
|20 May 2012||The competition starts! the competition data-sets are released, the submission site is opened for submitting your solutions.|
|30 Jun 2012||The competition ends, the results are announced.|
|14 Jul 2012:||The papers from the active participants are due.|
|5-8 Sep 2012||A special session takes place at ICGI 2012.|
|Winter 2012-2013||The winner is expected to submit a paper to the Machine Learning Journal special issue (see the CfP) that will follow the ICGI 2012 conference|
2 Different phases
In order to allow participants to train their systems before the official competition, a first phase opened in March 2012. At that time, you could download testing data that were not the ones of the competition, but were generated using the same process (they are still available at the end of the download page). The kind of machine that generate each sample is given in the name of each data set. The solution files, which contains the true probabilities for the test files and that were obtained using the target machines, are also available.
You were able to train your algorithm(s) on these data and to submit your results. You then received feedbacks as your score, the one of the baselines, and your rank between all submission to this problem was shown. Some baseline algorithm are also available so that you can compare your approach with them.
The real competition starts on May 20th, when the competition data sets were released. You then have access to the data sets of the competition. You also will be able to submit your results, but the feedback only consists in your rank among competitors for the given problem. The kind of machine that generates the set is not be available.