Recurrent Recommender Networks

By Chaoxia Wu, Amr Ahmed, Alex Beutel, Alex Smola, How Jing. 2017

758 citations recorded by the source corpus.

Recommender systems traditionally assume that user profiles and movie attributes are static. Temporal dynamics are purely reactive, that is, they are inferred after they are observed, e.g. after a user's taste has changed or based on hand-engineered temporal bias corrections for movies. We propose Recurrent Recommender Networks (RRN) that are able to predict future behavioral trajectories. This is achieved by endowing both users and movies with a Long Short-Term Memory (LSTM) autoregressive model that captures dynamics, in addition to a more traditional low-rank factorization. On multiple real-world datasets, our model offers excellent prediction accuracy and it is very compact, since we need not learn latent state but rather just the state transition function.

Subfields: Experimental and Cognitive Psychology, Information Systems, Management Science and Operations Research

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Experimental and Cognitive Psychology

In Information Systems

In Management Science and Operations Research

Citations in the corpus

Recurrent Neural Networks with Top-k Gains for Session-based Recommendations (2017, methodology): “RNNs have also been used in more standard user-item collaborative filtering settings where the aim is to model the evolution of the user and items factors [21],[6] where the results are less striking, with the proposed methods barely outperforming standard matrix factorization methods.”
DeepFM: A Factorization-Machine based Neural Network for CTR Prediction (2017, methodology): “Several deep learning models are proposed in recommendation tasks other than CTR prediction (e.g., [ Covington et al. , 2016; Salakhutdi-nov et al. , 2007; van den Oord et al. , 2013; Wu et al. , 2016; Zheng et al. , 2016; Wu et al. , 2017; Zheng et al. , 2017 ] ).”
Dynamical Isometry and a Mean Field Theory of RNNs: Gating Enables Signal Propagation in Recurrent Neural Networks (2018, background): “, 2013) and recommendation systems (Hidasi et al., 2015; Wu et al., 2017).”
Sequential Recommender Systems: Challenges, Progress and Prospects (2019, methodology): “Except for the basic RNN, longshort-term-memory (LSTM)- [21] and gated recurrent unit (GRU)-based [7] RNN have also been developed to capture the long-term dependencies in a sequence.”
Deep Learning Inference in Facebook Data Centers: Characterization, Performance Optimizations and Hardware Implications (2018, background): “Future Trends: • Model Exploration: recent studies explore explicitly incorporating time into the event-probability models [7, 73].”
Neural Networks and Deep Learning (2018, neutral): “Several solutions [465, 534, 565] have been proposed in recent years for temporal modeling of recommender systems.”
Relational Representation Learning for Dynamic (Knowledge) Graphs: A Survey (2019, background): “In an abstract dynamic recommender-system problem, we have a set of users, a set of items, and a set of timestamped interactions between users and items, and we seek to recommend items to users based on their current tastes [234, 128].”
Latent Relational Metric Learning via Memory-based Attention for Collaborative Ranking (2017, background): “mmendation. Intherecentyearswe can easily observe the emerging numbers of neural network models that have been designed for a diverse range of recommendation tasks. Notably, Recurrent neural networks [38] and convolutional neural networks [30] have been exploited for sequence aware recommendations. ”
Self-Attentive Sequential Recommendation (2018, background): “We also don’t include temporal recommendation methods like TimeSVD++ [16] and RRN [17], which differ in setting from what we consider here.”
A review on deep learning for recommender systems: challenges and remedies (2018, neutral): “Semantic meanings of textual information extracted with CNNs are also utilized in recommender systems especially for context-aware recommender systems to provide more qualified recommendations (Wu et al. 2017b).”
Translation-based Recommendation (2017, methodology): “Recently, Recurrent Neural Networks (RNN) are introduced into recommender systems to capture temporal dynamics [Hidasi et al. , 2016; Wu et al. , 2017; Jing and Smola, 2017].”
Predicting Dynamic Embedding Trajectory in Temporal Interaction Networks (2019, methodology): “We compare JODIE with six state-of-the-art algorithms spanning three algorithmic categories: (1) Deep recurrent recommender models: in this category, wecomparewithRRN[45],LatentCross[8],Time-LSTM[52], andstandardLSTM.”
BERT4Rec: Sequential Recommendation with Bidirectional Encoder Representations from Transformer (2019, background): “, Recurrent Neural Network (RNN), for sequential recommendation and obtain promising results [7, 14, 15, 56, 59].”
DÏoT: A Federated Self-learning Anomaly Detection System for IoT (2018, methodology): “GRUs provide similar accuracy as other RNN approaches but are computationally less expensive [17], [18].”
Neural Rating Regression with Abstractive Tips Generation for Recommendation (2017, background): “ons to find better latent structures from the user interactions for improving the performance of collaborative filtering. To model the temporal dynamic information in the user interactions, Wu et al. [43] propose a recurrent recommender network which is able to predict future behavioral trajectories”
Time Interval Aware Self-Attention for Sequential Recommendation (2020, background): “Temporal recommendation [17, 29, 31, 35] focuses on modeling absolute timestamps to capture the temporal dynamics of users and items.”
Improving Sequential Recommendation with Knowledge-Enhanced Memory Networks (2018, methodology): “With the revival of neural networks, many studies try to adapt powerful sequential neural models, i.e., Recurrent Neural Networks (RNN), to sequential recommendation [35], including session-based RNN [15], user-based RNN [5] and attention-based RNN [18].”
Fairness in Recommendation Ranking through Pairwise Comparisons (2019, background): “network techniques to improve recommender accuracy [25, 16, 36]. Building real-world recommenders face a variety of challenges. Two that relate to the challenges in fairness are the temporal dynamics [33, 48, 26, 9] and biased training data [29, 15, 3]. These issues do not just make training dicult ”