RecVAE: A New Variational Autoencoder for Top-N Recommendations with Implicit Feedback

Shenbin, Ilya; Alekseev, Anton; Tutubalina, Elena; Malykh, Valentin; Nikolenko, Sergey I.

doi:10.1145/3336191.3371831

Cited by 150 publications

(98 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The paper [23] was among the first works incorporating standard autoencoder frameworks to collaborative filtering. More recent architectures such as MultVAE [15] and RecVAE [24] stepped further by moving towards variational autoencoders and introducing specific loss functions tailored to the task of collaborative filtering.…”

Section: Related Workmentioning

confidence: 99%

Performance of Hyperbolic Geometry Models on Top-N Recommendation Tasks

Mirvakhabova

Frolov

Khrulkov

et al. 2020

Fourteenth ACM Conference on Recommender Systems

View full text Add to dashboard Cite

We introduce a simple autoencoder based on hyperbolic geometry for solving standard collaborative filtering problem. In contrast to many modern deep learning techniques, we build our solution using only a single hidden layer. Remarkably, even with such a minimalistic approach, we not only outperform the Euclidean counterpart but also achieve a competitive performance with respect to the current state-of-the-art. We additionally explore the effects of space curvature on the quality of hyperbolic models and propose an efficient data-driven method for estimating its optimal value. CCS Concepts: • Information systems → Recommender systems; Collaborative filtering; • Mathematics of computing → Geometric topology; • Computing methodologies → Neural networks.

show abstract

Section: Related Workmentioning

confidence: 99%

Performance of Hyperbolic Geometry Models on Top-N Recommendation Tasks

Mirvakhabova

Frolov

Khrulkov

et al. 2020

Fourteenth ACM Conference on Recommender Systems

View full text Add to dashboard Cite

show abstract

“…Thus, VAEs are generative models that can generate new data from the original dataset. Originally, VAEs and autoencoders are designed for image processing tasks such as image denoising [28], but the studies on these models have spread to various domains, including document modeling [29], [30] and also behavior recommendation [31], [32]. The application of VAEs in recommendation problem is expected because the main idea of this model is related to lower-dimensional representation, a technique that have been employed in many recommendation methods, such as matrix factorization [23], probabilistic matrix factorization [25], non-negative factorization [24].…”

Section: Related Workmentioning

confidence: 99%

“…These advantages make VAEs more useful and attractive in real-world applications. Some studies [32], [33] extended Mult-VAE and make some changes, such as new prior, new architecture, new training procedure, to improve the performance. We also notice that autoencoders are also applied to recommendation problems, which can be found in [34].…”

Section: Related Workmentioning

confidence: 99%

“…A solution for this problem is adjusting the regularization term of the objective function in VAEs, which has been shared by [31], [35]. Another approach is changing the training procedure for VAEs, which has been suggested in [32], [36]. Because updating the encoder is usually more complex and challenging than updating the decoder, it requires more epochs to achieve a good performance.…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

A Variational Autoencoder Mixture Model for Online Behavior Recommendation

Nguyen

Cho

2020

IEEE Access

View full text Add to dashboard Cite

Online behavior recommendation is an attractive research topic related to social media mining. This topic focuses on suggesting suitable behaviors for users in online platforms, including music listening, video watching, e-commerce, to name but a few to improve the user experience, an essential factor for the success of online services. A successful online behavior recommendation system should have the ability to predict behaviors that users used to performs and also suggest behaviors that users never performed before. In this paper, we develop a mixture model that contains two components to address this problem. The first component is the user-specific preference component that represents the habits of users based on their behavior history. The second component is the latent group preference component based on variational autoencoder, a deep generative neural network. This component corresponds to the hidden interests of users and allows us to discover the unseen behavior of users. We conduct experiments on various real-world datasets with different characteristics to show the performance of our model in different situations. The result indicates that our proposed model outperforms the previous mixture models for recommendation problem. INDEX TERMS Online behavior recommendation, Mixture model, Variational autoencoder I. INTRODUCTION Recommending online user behavior is an essential component in many online platforms to improve the user experience. These platforms aim to predict behaviors such as listening to a song, watching a video, purchasing a product, that users are more likely to perform in the future and then suggest that behaviors to users. We can also understand behavior recommendation as suggesting consumed items for users. Thus, in this paper, we may use both term "behavior" and "item" with the same meaning.

show abstract

“…In the field of recommender systems, although providing explanations is not a necessity, it has been widely shown to improve users satisfaction and trust in the system [17,40,42]. The use of neural networks does not consistently lead to an improvement in recommendation [8], but their ubiquity across various recent methods [14,15,18,27,28,37,44] brings us to the study of neural networks interpretability.…”

Section: Introductionmentioning

confidence: 99%

Making Neural Networks Interpretable with Attribution: Application to Implicit Signals Prediction

Afchar¹,

Hennequin²

2020

Fourteenth ACM Conference on Recommender Systems

View full text Add to dashboard Cite

Explaining recommendations enables users to understand whether recommended items are relevant to their needs and has been shown to increase their trust in the system. More generally, if designing explainable machine learning models is key to check the sanity and robustness of a decision process and improve their efficiency, it however remains a challenge for complex architectures, especially deep neural networks that are often deemed "black-box". In this paper, we propose a novel formulation of interpretable deep neural networks for the attribution task. Differently to popular post-hoc methods, our approach is interpretable by design. Using masked weights, hidden features can be deeply attributed, split into several input-restricted sub-networks and trained as a boosted mixture of experts. Experimental results on synthetic data and real-world recommendation tasks demonstrate that our method enables to build models achieving close predictive performances to their non-interpretable counterparts, while providing informative attribution interpretations.

show abstract

RecVAE: A New Variational Autoencoder for Top-N Recommendations with Implicit Feedback

Cited by 150 publications

References 21 publications

Performance of Hyperbolic Geometry Models on Top-N Recommendation Tasks

Performance of Hyperbolic Geometry Models on Top-N Recommendation Tasks

A Variational Autoencoder Mixture Model for Online Behavior Recommendation

Making Neural Networks Interpretable with Attribution: Application to Implicit Signals Prediction

Contact Info

Product

Resources

About