This challenge requires to label input data with a certain class, based on the properties of the data.

There are usually two types of challenges:

  • challenges which require to classify correctly (usually with an acceptable error threshold) the test cases given, in the minimal amount of bytes.

  • challenges which require to classify as well as possible the chosen test cases.

For challenges, it is advised to prepare three test cases sets: a training, a validation and a test set. The training set can be used by participants to train machine learning based approaches. The validation set can be used by participants to test the performance of their algorithm. The test set should not be given to participants and is used by the author of the challenge to actually evaluate the score of each answer.

This way, there is no possibility for participants to optimize their answer for the test set, and they thus are forced to make a legitimate attempt at identifying the discriminatory features of the data using the training and validation sets.