基于稳定学习的多兴趣序列推荐网络OA北大核心CSTPCD

Multi-interest models,which extract interests of a user as multiple representation vectors,have shown promising perfor-mances for sequential recommendation.However,considering that multiple interests of a user are usually highly correlated,the mod-el has chance to learn spurious correlations between noisy interests and target items.Once the data distribution changes,the correla-tions among interests may also change,and the spurious correlations will mislead the model to make wrong predictions.To solve such problem,we propose a novel model of Multi-Interest network with Stable Learning(MISL),which attempts to de-correlate the extracted interests,and thus spurious correlations can be eliminated.MISL applies an attentive module to extract multiple interests,and then selects the most important one for making final predictions.Meanwhile,MISL incorporates a weighted correlation estima-tion loss based on independence criterion,with which training samples are weighted,to minimize the correlations among extracted interests.Extensive experiments have been conducted under both Out-of-Distribution(OOD)and random settings,and up to 36.8%and 21.7%relative improvements are achieved respectively.