Metric Learning for Special Settings

Abstract

In the previous two chapters, we have reviewed methods to learn metrics in a standard supervised classification setup. In this chapter, we present extensions of metric learning to more complex settings that require specific formulations. Section 6.1 deals with multi-task and transfer learning, while Section 6.2 is devoted to learning a metric for ranking. Section 6.3 addresses the problem of learning a metric from semi-supervised data or in a domain adaptation setting. Finally, Section 6.4 presents approaches to learn metrics specifically for histogram data.