Abstract
For students and course participants, the website challengedata.ens.fr offers new supervised learning data processing challenges for the 2025 season. These challenges are proposed by companies or scientists and are based on real-world problems they encounter in their work. They are designed to foster scientific exchange through the sharing of data and algorithms.
Each challenge provides labeled data as well as test data. Participants submit their predictions, calculated using the test data, via the website. The site calculates a score using a specified error metric. It provides participants with a ranking, allowing them to evaluate their results within a broad community. The challenges begin on January 8, 2025. An interim deadline will take place in March, featuring an evaluation of predictions based on new test data. The final deadline will be in December 2026.
This year, the challenges were organized and supervised at the ENS and the Institut Louis Bachelier by Enzo Dalby, with the participation of Mohamed Fahmaoui, Martin Recroix, Tiphanie Motel, Nathanael Cuvelle-Magar, and Etienne Lempereur. The organization of these data challenges is supported by the CFM Chair at the École normale supérieure. The following 11 challenges were organized and presented during the first two seminar sessions:
- Learning radiological and oncological anatomy using few-shot learning (Raidium);
- Identification of toxic gases (Bertin Technologies);
- AssurPrime: Can You Predict the Insurance Premium? (Crédit Agricole Assurances);
- Predict parking violations! (Egis);
- Behavior-based job recommendations using online machine learning and collaborative filtering (HrFlow.ai);
- Parkinson’s disease: predicting and correcting biases in motor score assessment (Institut du Cerveau);
- Predicting overall survival in patients with myeloid leukemia (QRT);
- Medical segmentation of anatomical structures with missing labels (Raidium);
- EchoCem: Assessment of cement quality through ultrasound image segmentation (SLB);
- Real-time prediction of platform wait times (SNCF-Transilien);
- Improvement of industrial quality control using computer vision (Valeo).