Transfer Learning in Complex Environments

Abstract

The application environment is dynamically changing, so does the algorithm. In order to cope with the changing environment, there are several new transfer learning algorithms being developed. We define the complex environment mainly based on the traits of the training data since data is the key to modern transfer learning. In this chapter, we briefly introduce some of these complex environments and show how transfer learning algorithms can be adapted to deal with such new challenges. Note that there are multiple new settings in recent literature and we cannot cover them all.