Background

Multi-label learning deals with the problem where each example is associated with multiple labels and thus encompasses traditional supervised learning (single-label) as its special case.

Though methods for learning from multi-label textual data have been proposed since 1999, the recent years have witnessed an increasing number and diversity of applications, such as image/video annotation, bioinformatics, web search and mining, music categorization, collaborative tagging and directed marketing.

Learning from multi-label data stretches across several aspects of supervised learning tasks, including classification, ranking, semi-supervised learning, active learning and dimensionality reduction, and across several learning paradigms, such as decision trees, nearest neighbor classifiers, neural networks, ensemble methods, support vector machines, kernel methods, genetic algorithms, etc.

It poses several old and new research challenges, such as exploiting label correlation to improve predictive performance, exploiting structure and semantic relationships among the labels to improve predictive performance and computational efficiency, and scaling learning methods to very large number of labels and examples. In addition, multi-label learning is closely related to other learning frameworks, such as the newly proposed multi-instance multi-label learning (MIML).

Aims and Scope

The goal of this workshop is to bring researchers and practicioners that work on various aspects of multi-label learning into a fruitful dicussion about the state-of-the-art and the remaining open problems, and to offer them an opportunity to identify new promising research directions. To achieve this goal we are soliciting two types of contributions: a) mature research results, and b) interesting preliminary results or stimulating position statements. In addition, the workshop will feature at least one discussion session to allow for a more interactive and engaging experience.

Topics of Interest

The workshop's topics of interest include (but are not limited to):

• Classification of multi-label data
• Ranking of multi-label data
• Statistical characterizations of multi-label data sets
• Visualization of multi-label data sets
• Evaluation metrics for multi-label learning methods
• Exploiting label structure and relationships (trees, ontologies, etc)
• Learning label structure and relationships
• Learning from multiple continuous target variables
  
• Online learning from multi-label data
  
• Hierarchical multi-label classification and ranking
  
• Dimensionality reduction of multi-label data
  
• Clustering multi-label data
  
• Semi-supervised learning from multi-label data
  
• Learning association rules from multi-label data
  
• Scalable methods for learning with very large number of labels
  
• Multi-instance multi-label learning
  
• Active learning from multi-label data
  
• Applications of multi-label learning in bioinformatics
  
• Semantics annotation of images and video
  
• Multi-label learning from music
• Automated tag recommendation in collaborative tagging systems