Abstract:Predicting potential user behaviors is of great importance in recommendation systems. Existing behavior prediction works mainly aim to construct the user behavior preference based on a large amount of user interaction data, but the prediction performance deteriorates when facing the users with a few interactions. A common and effective approach to alleviate the interaction-sparse problem is to incorporate the auxiliary information as much as possible, in which the hybrid (Variational Autoencoder) VAE method reports the optimal performance with the advantages of non-linear modeling and comprehensive integration. However, the hybrid VAE only focuses on the attribute data, which limits the richness and robustness of the user preference representation in interaction-sparse scenario. In this paper, we propose a novel hybrid VAE model, named Hi erarchical C onstrained V ariational A uto E ncoder Jointly Attributes and Neighbors (HiCVAE), to tackle the above limitations. Specifically, HiCVAE additionally introduces fine-grained neighbors' behaviors as auxiliary information to enhance the richness and robustness of the user preference representation on the basis of the attribute data, by learning a hierarchical prior instead of a standard Gaussian prior. The prior on the attribute data is responsible for the community preference relevant to a group with the same attributes. Further, the conditional prior over the neighborhood data can capture the personalized information and provide further constraints for the new user preference representation. In addition, a Preference Consistency Loss is proposed to restrict the reconstructed interest to be consistent with the user's real interest and eliminate the irrelevant information on the neighborhood data. Thorough and extensive experiments conducted on two real-world datasets demonstrate that the HiCVAE significantly outperforms the state-of-the-art methods in various interaction-sparse scenarios (sparse, highly sparse and extremely sparse).

Implicit Relation-Aware Social Recommendation with Variational Auto-Encoder

Towards Context-aware Social Recommendation Via Individual Trust

Disentangled Variational Autoencoder for Social Recommendation

Dual Variational Graph Reconstruction Learning for Social Recommendation

Hierarchical Constrained Variational Autoencoder for interaction-sparse recommendations

Exploiting Structural and Temporal Influence for Dynamic Social-Aware Recommendation

A recommender system fused with implicit social information through network representation learning

VIGA: A variational graph autoencoder model to infer user interest representations for recommendation

DualVAE: Dual Disentangled Variational AutoEncoder for Recommendation

A Trust Networks Recommender Algorithm Based On Latent Factor Model

Fusion of Personalized Implicit Relations for Social Recommendation

A Survey on Variational Autoencoders in Recommender Systems

Network Representation Learning Enhanced Recommendation Algorithm

A Collective Variational Autoencoder for Top-$N$ Recommendation with Side Information

Social Recommendation With Learning Personal and Social Latent Factors

Disentangled-feature and composite-prior VAE on social recommendation for new users

Collaborative Filtering Recommendation Algorithm Based on Variational Inference

Variational Bandwidth Auto-encoder for Hybrid Recommender Systems

VAEGAN: A Collaborative Filtering Framework based on Adversarial Variational Autoencoders

Self-supervised variational autoencoder towards recommendation by nested contrastive learning

IDVT: Interest-aware Denoising and View-guided Tuning for Social Recommendation