Data Challenge Presentation: Cross-Domain MetaDL

The new Cross-Domain MetaDL challenge is part of the ChaLearn meta-learning series. It has a special league for New in ML participants (with prizes and certificates) and a detailed tutorial with no prerequisites (i.e., no previous meta-learning knowledge required). The competition is part of the NeurIPS'22 program, and the winners will be invited to co-author the analysis paper with the organizers to appear in PMLR. The focus is on "cross-domain" meta-learning, aiming at leveraging experience from previous tasks to solve new tasks efficiently. While our previous challenge addressed within-domain few-shot learning for N-way k-shot tasks (i.e., N class classification problems with k training examples), this challenge proposes "any-way" and "any-shot" tasks drawn from various domains (healthcare, ecology, biology, manufacturing, and others), chosen for their humanitarian and societal impact. Code submissions will be blind-tested on CodaLab, and the winners' code will be open-sourced.