Active Learning of Label Ranking Functions
Klaus Brinker - University of Paderborn
The effort necessary to construct labeled sets of examples in a supervisedlearning scenario is often disregarded, though in many applications, it is atime-consuming and expensive procedure. While this already constitutes a majorissue in classification learning, it becomes an even more serious problem whendealing with the more complex target domain of total orders over a set ofalternatives. Considering both the pairwise decomposition and the constraintclassification technique to represent label ranking functions, we introduce anovel generalization of pool-based active learning to address this problem.