Modern applications in areas such as information retrieval, image analysis and computational biology have triggered a need for specialized machine learning methods that deliver joint predictions for multiple targets. Recently, new algorithms for multi-target prediction problems have been proposed in different subfields of machine learning, and most of the recent methods intend to exploit, in one way or another, dependencies between targets. The goal of this project is to elaborate on the idea of exploiting target dependence, thereby contributing to a better understanding of existing multi-target prediction methods.