New Framework for Deep Mutual Studying (DML) to Enhance Multi-task Recommender Methods

This paper introduces Deep Mutual Studying (DML) to enhance multi-task recommender techniques. It integrates Cross Process Characteristic Mining (CTFM) to share task-specific inputs and International Data Distillation (GKD) to refine prediction consistency. DML enhances information switch and overcomes process conflicts, validated by offline experiments and on-line A/B exams.  Learn Extra

Leave a Reply

Your email address will not be published. Required fields are marked *