Many classification problems may be difficult to formulate in the traditional supervised setting, where both training and test samples are individual feature vectors. It may be the case that samples are better described by sets of feature vectors, that labels are only available for sets rather than individual samples, or, if individual labels are available, that these are not independent. To better deal with such problems, several extensions of supervised learning have been proposed, where either training and/or test objects are sets of feature vectors. However, such extensions are often proposed independently, disregarding important similarities and differences with other existing classification problems. In this work, we provide an overview of such learning scenarios, propose a taxonomy to illustrate the relationships between them, and discuss directions for further research in these areas.
View on arXiv