Translational Recommender Networks

Œis paper proposes a new neural architecture for collaborative ranking with implicit feedback. Our model, LRML (Latent Relational Metric Learning) is a novel metric learning approach for recommendation. More speci€cally, instead of simple push-pull mechanisms between user and item pairs, we propose to learn latent relations that describe each user item interaction. Œis helps to alleviate the potential geometric inƒexibility of existing metric learing approaches. Œis enables not only beŠer performance but also a greater extent of modeling capability, allowing our model to scale to a larger number of interactions. In order to do so, we employ a augmented memory module and learn to aŠend over these memory blocks to construct latent relations. Œe memory-based aŠention module is controlled by the user-item interaction, making the learned relation vector speci€c to each user-item pair. Hence, this can be interpreted as learning an exclusive and optimal relational translation for each user-item interaction. Œe proposed architecture demonstrates the state-of-the-art performance across multiple recommendation benchmarks. LRML outperforms other metric learning models by 6% − 7.5% in terms of Hits@10 and nDCG@10 on large datasets such as Netƒix and MovieLens20M. Moreover, qualitative studies also demonstrate evidence that our proposed model is able to infer and encode explicit sentiment, temporal and aŠribute information despite being only trained on implicit feedback. As such, this ascertains the ability of LRML to uncover hidden relational structure within implicit datasets.