Recommender systems have become an important instrument to connect people to information. Sparse, complex, and rapidly growing data presents new challenges to traditional recommendation algorithms. To overcome these challenges, various deep learning-based recommendation algorithms have been proposed. Among these, Variational AutoEncoder (VAE) -based recommendation methods stand out. VAEs are based on a flexible probabilistic framework, which is robust for data sparsity and compatible with other deep learning-based models for dealing with multimodal data. In addition, the deep generative structure of VAEs helps to perform Bayesian inference in an efficient manner. VAE-based recommendation algorithms have given rise to many novel graphical models, and they have achieved promising performance. In this article, we conduct a survey to systematically summarize recent VAE-based recommendation algorithms. Four frequently used characteristics of VAE-based recommendation algorithms are summarized, and a taxonomy of VAE-based recommendation algorithms is proposed. We also identify future research directions for, advanced perspectives on, and the application of VAEs in recommendation algorithms, to inspire future work on VAEs for recommender systems.

Hashing is a practical approach for the approximate nearest neighbor search. Deep hashing methods, which train deep networks to generate compact and similarity-preserving binary codes for entities (e.g. images), have received lots of attention in the information retrieval community. A representative stream of deep hashing methods is triplet-based hashing that learns hashing models from triplets of data. The existing triplet-based hashing methods only consider triplets that are in the form of $(q,q^{+},q^{-})$ , where $q$ and $q^{+}$ are in the same class and $q$ and $q^{-}$ are in different classes. However, the number of possible triplets is approximately the cube of training examples, triplets used in the existing methods are only a small fraction of all possible triplets. This motivates us to develop a new triplet-based hashing method that adopts many more triplets in training phase. We propose Deep Listwise Triplet Hashing (DLTH) that introduces more triplets into batch-based training and a novel listwise triplet loss to capture the relative similarity in new triplets. This method has a pipeline of two steps. In Step 1, we propose a novel way to generate triplets from the soft class labels obtained by knowledge distillation module, where the triplets in the form of $(q,q^{+},q^{-})$ are a subset of the newly obtained triplets. In Step 2, we develop a novel listwise triplet loss to train the hashing network, which seeks to capture the relative similarity between images in triplets according to soft labels. We conduct comprehensive image retrieval experiments on four benchmark datasets. The experimental results show that the proposed method has superior performances over state-of-the-art baselines.

Item recommendation has become a significant means to help people discover interesting items. Meanwhile, plenty of reviews and ratings in recommender system can be utilized to relieve data sparsity problem. However, existing review-based approaches ignore the influence of static preference of user and the static characteristics of item, which could reflect long-term and stationary property, and guide feature extraction from reviews. Moreover, adaptive property, i.e., the importance of the historical records to each user and item, is not fully exploited in previous works. In this paper, we propose an Attentionbased Adaptive Memory Network (AAMN) model to leverage historical reviews and ratings systemically. Specifically, we propose an attention mechanism guided by the static features to learn the importance of different historical records, for modeling the adaptive features of users and items. Notably, this paper is the first to bring static features into adaptively extracting semantic information from reviews, which can not only characterize user and item from a global view, but also assist to distinguish the importance of different reviews. In addition to the attention mechanism, we propose a non-linear feature fusing layer and a deep interaction layer to combine the static features and adaptive features, which capture underlying interactions among these features. To further improve prediction accuracy and training efficiency, we propose a dynamic sampling strategy for model training. We conduct extensive experiments on 16 benchmark datasets from Amazon and Yelp. The results demonstrate that our model outperforms the state-of-the-art models.

Due to the easy access, implicit feedback is often used for recommender systems. Compared with point-wise learning and pair-wise learning methods, list-wise rank learning methods have superior performance for top- \(N\) recommendation. Recent solutions, especially the list-wise methods, simply treat all interacted items of a user as equally important positives and annotate all no-interaction items of a user as negatives. For the list-wise approaches, we argue that this annotation scheme of implicit feedback is over-simplified due to the sparsity and missing fine-grained labels of the feedback data. To overcome this issue, we revisit the so-called positive and negative samples. First, considering the loss function of list-wise ranking, we analyze the impact of false positives and negatives theoretically. Second, based on the observation, we propose a self-adjusting credibility weight mechanism to re-weigh the positive samples and exploit the higher-order relation based on item–item matrix to sample the critical negative samples. In order to prevent the introduction of noise, we design a pruning strategy for critical negatives. Besides, to combine the reconstruction loss function for the positive samples and critical negative samples, we develop a simple yet effective VAEs framework with linear structure, which abandons the complex non-linear structure. Extensive experiments are conducted on six public real-world datasets. The results demonstrate that, our VAE* outperforms other VAE-based models by a large margin. Besides, we also verify the effect of denoising positives and exploring critical negatives by ablation study.