Neural Social Recommendation With User Embedding

Abstract

Social information is usually jointly utilized with rating information to help the traditional recommendation system providing more personalized services, while how to make full use of social information to build better recommendation models still faces lots of challenges. In this paper, we propose a novel social recommendation model taking advantage of both deep and shallow model, with deep autoencoders acting as the nonlinear feature extractor and MF-based method being used to depict the user's preferences. Then, motivated by the idea of word2vec, we provide an appealing method that embeds users into latent space and meanwhile preserves the structural information of social networks during the embedding. Also, the embedded latent features of each user are corresponding to the dual roles the user plays in the recommendation. Furthermore, we design a loss function for the holistic training of the model, and our loss function is made up mainly of three parts which embody the effects of different factors on the rating predictions. Specifically, the loss function (i) captures the personal preference from user-item adoption matrix based on matrix factorization, (ii) discriminates the two different social functions of users and further evaluates the effects of interpersonal influence through user embedding and social influence matrix, (iii) and avoids overfitting by imposing a quadratic regularization penalty. As a result, our model can predict the missing ratings with the MF-based method by consuming the latent features of users and items extracted by the deep model. The experiments show that our method outperforms existing methods and performs well on cold start users. Social recommendation, user embedding, dual role, auto-encoder. INDEX TERMS

Deep Learning-Based Recommendation System: Systematic Review and Classification

“…Optimizing recommendation models involves diverse techniques, including knowledge distillation [224] that transfers knowledge between models. Loss functions, like the one designed by Xia et al [270] for holistic training and model compression approaches [271], enhance model efficiency. Evaluation techniques such as performance evaluation [272] and quantile loss [225] provide insights into model effectiveness.…”

“…These techniques include knowledge distillation, where the work of Pan et al [224] developed two novel heterogeneous knowledge distillation methods, namely feature-level, and label-level, to build relations between IAE and UAE models. Additionally, the category includes loss functions, such as the one designed by Xia et al [270] for the holistic training of the model, which consists of three parts representing the effects of different factors on rating predictions and model compression as outlined in the work of Zhao et al [271]. These techniques aim to enhance the performance of recommendation systems by simplifying the models and improving their efficiency.…”

Deep transfer learning with multimodal embedding to tackle cold-start and sparsity issues in recommendation system

“…SimilarMF, proposed in [ 42 ] used embedding and social information along with a rating matrix to produce user-user and item-item similarity matrices which showed improved results. Neural Social Recommendation [ 43 ], a deep model based on matrix factorization, makes use of social information with user embedding to exploit user, item latent features for improving the prediction process. To avoid sparsity in the graph, the Collaborative Similarity Embedding (CSE) technique [ 44 ] leveraged direct relations from the input graph to discover similarity matrices.…”