Collaborative Recommendation with Auxiliary Data: A Transfer Learning View

Intelligent recommendation technology has been playing an increasingly important role in various industry applications such as e-commerce product promotion and Internet advertisement display. Besides users' feedbacks (e.g., numerical ratings) on items as usually exploited by some typical recommendation algorithms, there are often some additional data such as users' social circles and other behaviors. Such auxiliary data are usually related to users' preferences on items behind the numerical ratings. Collaborative recommendation with auxiliary data (CRAD) aims to leverage such additional information so as to improve the personalization services, which have received much attention from both researchers and practitioners. Transfer learning (TL) is proposed to extract and transfer knowledge from some auxiliary data in order to assist the learning task on some target data. In this paper, we consider the CRAD problem from a transfer learning view, especially on how to achieve knowledge transfer from some auxiliary data. First, we give a formal definition of transfer learning for CRAD (TL-CRAD). Second, we extend the existing categorization of TL techniques (i.e., adaptive, collective and integrative knowledge transfer algorithm styles) with three knowledge transfer strategies (i.e., prediction rule, regularization and constraint). Third, we propose a novel generic knowledge transfer framework for TL-CRAD. Fourth, we describe some representative works of each specific knowledge transfer strategy of each algorithm style in detail, which are expected to inspire further works. Finally, we conclude the paper with some summary discussions and several future directions.