Meta-Learned User Preference Estimator with Attention Network for Cold-Start Recommendation

Abstract

Abstract One crucial challenge in the recommendation research field is the cold-start problem. Meta-learning is a feasible algorithm to reduce the error of cold-start recommendation because it can adjust to new tasks rapidly through relatively few updates. However, meta-learning does not take the diverse interests of users into account, which limits the performance improvement in cold-start scenarios. We proposed a recommendation model called attentional meta-learned user preference estimator that combines attention network and meta-learning. This method enhances the ability of modelling the personalized user interest by learning the weights between users and items based on attention mechanism, then improves the performance of cold-start recommendation. We validated the model with two publicly available datasets in the recommendation field. Compared with the three benchmark methods, the proposed model reduces the mean absolute error by at least 2.3% and the root mean square error of 2.5%.