The aim of the competition is to identify the most successful technique for learning state machines from samples of their behaviour. In other words, given a set of strings that are classified as either positive (i.e. they are accepted by the target machine) or negative (they are rejected), the learned model should accurately classify new strings as positive and negative.
Participation in the competition consists in attempting to solve cells. A cell is considered to be solved if all of the 5 problems it contains are solved with a BCR score of at least 0.99 (see the protocol page). The first technique to solve a hardest cell, among those solved in the competition, will be the winner. The organizers will add additional problems if all problems appear to be solved more than one month before the final deadline.
The competition process is illustrated on the image below. Click on the different nodes to have detailed information