Multi-Label Learning

This workshop aims to bring researchers and practitioners that work on different aspects of multi-label learning into a fruitful discussion about the state-of-the-art, remaining open problems, and new promising research directions. The workshop welcomes both theoretical multi-label learning papers, as well as papers discussing real-world applications involving multi-label data. We are particularly interested in submissions of papers at the intersection of human-centered trustworthy machine learning aspects (e.g., interpretability, explainability, bias, fairness, safety, active learning) and multi-label learning, as well as papers discussing persistent challenges in multi-label data (e.g., data streams, extreme classification, class imbalance, missing labels, noisy labels, real-world industrial aspects). The workshop is open to papers dealing with any type of data modality (e.g., textual, visual, tabular). The best paper of the workshop will be awarded with a free ECML PKDD 2022 registration.

Topics

• Methods for multi-label classification and ranking
• Statistical and semantic analysis of multi-label data
• Evaluation measures and strategies for multi-label data
• Learning label structure and relationships
• Embeddings and representations of multi-label data
• FAIR multi-label data
• Explainable multi-label learning
• Trustworthy multi-label learning
• Extreme multi-label classification
• Deep multi-label learning
• Dealing with class imbalance
• Learning from partially labeled data
• Weak multi-label learning
• Zero-shot and few shot multi-label learning
• Semi-supervised learning from multi-label data
• Active learning from multi-label data
• Dimensionality reduction (feature selection) of multi-label data
• Hierarchical multi-label classification and learning
• Learning from streams of multi-label data
• Applications in a variety of domains, including life sciences, medicine, engineering, and multimedia

Form

The workshop will have a half-day duration, featuring invited talks from world-class experts, oral presentations of all accepted papers, a poster session for additional discussion of each accepted paper among workshop participants, and finally a session dedicated to openly discussing challenges and opportunities in multi-label learning. The workshop will take place as a hybrid event. However, as interactions and discussions are much easier and more fruitful face-to-face, we encourage especially speakers, but also participants, to join us physically in beautiful Grenoble.