Search papers, labs, and topics across Lattice.
This paper investigates zero-shot cross-domain knowledge distillation (KD) to improve ranking models in a low-traffic music recommendation system by transferring knowledge from a high-traffic video recommendation system. The study evaluates various KD techniques in offline and live experiments, demonstrating the effectiveness of this approach for multi-task ranking models. Results show that zero-shot cross-domain KD can significantly improve the performance of ranking models in data-scarce environments without requiring dedicated teacher model training.
You can boost ranking model performance in low-traffic recommendation systems by directly distilling knowledge from a large-scale, but different, domain like video recommendations.
Knowledge Distillation (KD) has been widely used to improve the quality of latency sensitive models serving live traffic. However, applying KD in production recommender systems with low traffic is challenging: the limited amount of data restricts the teacher model size, and the cost of training a large dedicated teacher may not be justified. Cross-domain KD offers a cost-effective alternative by leveraging a teacher from a data-rich source domain, but introduces unique technical difficulties, as the features, user interfaces, and prediction tasks can significantly differ. We present a case study of using zero-shot cross-domain KD for multi-task ranking models, transferring knowledge from a (100x) large-scale video recommendation platform (YouTube) to a music recommendation application with significantly lower traffic. We share offline and live experiment results and present findings evaluating different KD techniques in this setting across two ranking models on the music app. Our results demonstrate that zero-shot cross-domain KD is a practical and effective approach to improve the performance of ranking models on low traffic surfaces.
