Contest Rules

The winner is awarded with a cash prize of 500 euros from ab medica s.p.a.
It is possible to participate individually or in a group (max 3 participants). In case of group participation, however, the registration of all members is required by filling out the form and paying the fee (30 euros and 150 euros for M.Sc. and Ph.D. students, respectively).
For those who are already registered for the conference, participation in the contest is free of charge. However, formal registration for the competition is still required.
In case of victory by a group, the prize is divided equally among the participants.
Participants will be required to develop algorithms that can maximize
- within-subject accuracy using data from the Motor Imagery dataset (dataset A), in the context of a binary classification problem (right hand and left hand)
- cross-subject accuracy using data from the Emotion dataset (dataset B), within two binary problems (high and low valence; high and low arousal).
Both datasets can be downloaded at the page that automatically opens after filling out the entry form.
On the day of the contest, set for Friday, Oct. 28 at 9 a.m., applicants will receive new data, without labels, on which they will test their algorithms via their laptop. In particular, (i) a new session for each of the seven subjects in dataset A and (ii) data from 3 new subjects for dataset B will be furnished.
The outcome of the classification test will be compared with the labels held by the panel of judges. Two hours after the start of the contest, a ranking order will be assigned to the participants based on the highest classification accuracy achieved on the new data
For each of the three binary classification problems, participants are given a score equal to the position achieved in the ranking order. In case of ex-aequo, the same score is given. The final score is calculated by summing the scores achieved in the three classification problems. The participant with the lowest total score wins. In case of a tie, the participant with the highest average accuracy wins. In case of a further tie, a draw will be made.
On the day of the Contest, participants are allowed to conduct additional learning steps with unlabeled data before taking the classification test.
For dataset B, it is permissible for participants to use the label class "y_IAPS_x" and the label class "y_SAM_x" (see the presentation of EMOTION DATASET downloadable at the same page of datasets) alternately or in a combination. The type of use of the two labels must be communicated to the jury before the start of the competition.
The day of the Contest, participants must produce

a .txt file containing the outcome of three binary classification problems expressed in terms of sequences of "1" and "2" values. One binary classification problem for dataset A. Two binary classification problems for dataset B.
a Matlab script capable of generating labels of type y_MIX_x from the labels y_IAPS_x" and "y_SAM_x", in the case of adopting an algorithm that combines the use of labels y_IAPS_x" and "y_SAM_x" for dataset B.

The Winner acknowledges and accepts that all intellectual property rights and economic and industrial exploitation rights relating to and/or in any way connected with the "Best Performer Algorithms" belong exclusively to ab medica, which may therefore, at its sole discretion, proceed with the filing of any patent application, at its sole care and expense. The Winner agrees, accordingly and as of now, to put in place and/or collaborate with ab medica to put in place all the procedures necessary to transfer and/or recognize to AB Medica any related right. This is without prejudice to the recognition of the Winner's moral right to be counted among the inventors, where the Algorithm is independently patentable.
The prize will be awarded only if the average accuracy threshold of 70% is exceeded.