Multi-label Learning by Simultaneously Exploiting Locality Underlying the Instance Space and the Label Space

Multi-label classification has attracted much attention in recent years due to various applications in real world. There have been many algorithms to deal with this problem. However, there is no algorithm that simultaneously exploits the locality in the instance space and label space which plays an important role in generating a satisfactory model. In this paper we present such an algorithm. It utilizes the locality underlying instance space and label space to regularize the learning process. Experiments on three distinct application domains validate the effectiveness of our proposed algorithm, and it achieves superior performance to some state-of-art algorithms.