Supervised learning algorithms require sufficient amount of labeled training data for learning robust prediction models. The field of Transfer Learning (TL) (also known as knowledge transfer) deals with utilizing knowledge from data-rich auxiliary domains to learn a reliable predictor for the domain of interest. This chapter presents a condensed review of the shallow TL literature (prior to the deep learning era). The chapter motivates the need for TL using an application. After an informal introduction to TL, a categorization of TL approaches based on the characteristics of the domains is presented. Next, the different transfer settings along with the challenges in each setting are described. The TL frameworks are delineated using a generic optimization problem. The chapter also discusses a few real-world applications used for benchmarking experiments for each transfer setting. Finally, the chapter concludes with some unexplored avenues in the TL research. © Springer Nature Switzerland AG 2020.