Parameter-Efficient Transfer from Sequential Behaviors for User Profiling and Recommendation

Inductive transfer learning has greatly impacted the computer vision and NLP domains, but existing approaches in recommender systems remain largely unexplored. Meanwhile, although there has been a large body of research making direct recommendations based on user behavior sequences, few of them attempt to represent and transfer these behaviors for downstream tasks. In this paper, we look in particular at the task of effectively learning a single user representation that can be applied to a diversity of tasks, from cross-domain recommendations to user profile predictions. Fine-tuning a large pre-trained network and adapting it to downstream tasks is an effective way to solve such an issue. However, fine-tuning is parameter inefficient considering that an entire model needs to be re-trained for every new task. To overcome this issue, we develop a parameter-efficient transfer learning architecture, termed as PeterRec, which can be configured on-the-fly to various downstream tasks. Specifically, PeterRec allows the pre-trained parameters unaltered during fine-tuning by injecting a series of re-learned neural networks, which are small but as expressive as learning the entire network. We perform extensive experimental ablation to show the effectiveness of learned user representation in five downstream tasks. Moreover, we show that PeterRec performs efficient transfer learning in multiple domains, where it achieves comparable or sometimes better performance relative to fine-tuning the entire model parameters.