Joint Learning Using Multiple Types of Data and Knowledge

Abstract

This chapter discusses joint learning research in biomedical domains. A brief review of the field of joint learning research is given, with emphases on the large-scale data and knowledge resources used for learning and the central biological questions involved. Two representative joint learning case studies are presented with algorithmic details. The two case studies involved two representative joint learning tasks, protein function classification and regulatory network learning, and two important algorithmic frameworks for joint learning, the kernel-based framework and probabilistic graphical models. A wide range of biological data and existing knowledge was also involved in these two studies.