weikangliang

Results 1 comments of weikangliang

@yuanninesuns 据我所知,这种训练方式确实有问题,就像MIND一样有信息泄露,使得训练不difficult。所以今年出了篇论文(User-Aware Multi-Interest Learning for Candidate Matching in Recommenders) https://dl.acm.org/doi/abs/10.1145/3477495.3532073,改进了这个问题,方法是要泄露大家都泄露,不只是target泄露!