Improved Recurrent Neural Networks for Session-based Recommendations
By Yong Kiam Tan, Xinxing Xu, Yong Liu. 2016
852 citations recorded by the source corpus.
Recurrent neural networks (RNNs) were recently proposed for the session-based recommendation task. The models showed promising improvements over traditional recommendation approaches. In this work, we further study RNN-based models for session-based recommendations. We propose the application of two techniques to improve model performance, namely, data augmentation, and a method to account for shifts in the input data distribution. We also empirically study the use of generalised distillation, and a novel alternative model that directly predicts item embeddings. Experiments on the RecSys Challenge 2015 dataset demonstrate relative improvements of 12.8% and 14.8% over previously reported results on the Recall@20 and Mean Reciprocal Rank@20 metrics respectively.
Subfields: Artificial Intelligence, Information Systems, Model Architectures, Recommender Systems
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Artificial Intelligence
- Xiangnan He
- Yongfeng Zhang
- Xu Chen
- Hongzhi Yin
- Yi Tay
- Julian McAuley
- Dietmar Jannach
- Naomie Binti Salim
In Information Systems
- Xiangnan He
- Yongfeng Zhang
- Xu Chen
- Hongzhi Yin
- Dietmar Jannach
- Naomie Binti Salim
- Aminu Da’u
- Malte Ludewig
In Model Architectures
- Dietmar Jannach
- Malte Ludewig
- Hongzhi Yin
- Xiangnan He
- Quan Z. Sheng
- Shoujin Wang
- Longbing Cao
- Yan Wang
In Recommender Systems
- Xiangnan He
- Yongfeng Zhang
- Dietmar Jannach
- Malte Ludewig
- Xu Chen
- Aminu Da’u
- Naomie Binti Salim
- Hongzhi Yin
Citations in the corpus
- Variational Autoencoders for Collaborative Filtering (2018, methodology): “it choice model [30]. It is also used in the cross-entropy loss2 for multi-class classification. For example, it has been used in recurrent neural networks for session-based sequential recommendation [8, 15, 16, 42, 45] and in feedward neural networks applied to Youtube recommendation [9]. The multi”
- Recurrent Latent Variable Networks for Session-Based Recommendation (2017, methodology): “we also cite the performance of recently proposed state-of-the-art alternatives in the same benchmark, namely the GRU-based method presented in [13], and the M2 and M4 approach introduced in [31].”
- Recurrent Neural Networks with Top-k Gains for Session-based Recommendations (2017, methodology): “[19] proposed data augmentation techniques to improve the performance of the RNN for session-based recommendations, however these techniques have the side effect of increasing training times as a single session is split into several sub-sessions for training.”
- AutoEmb: Automated Embedding Dimensionality Search in Streaming Recommendations (2020, background): “Activation function д (·) varies according to different prediction tasks, such as Sigmoid function for app installation prediction (regression) [5], and Softmax for buy-or-not prediction (classification) [29].”
- Evaluation of session-based recommendation algorithms (2018, methodology): “Algorithm MRR@20 HR@20 COV@20 POP@20 MRR@10 HR@10 COV@10 POP@10 ar”
- Neural Attentive Session-based Recommendation (2017, methodology): “(The baseline method is Improved GRU-Rec [40].”
- A Survey on Session-based Recommender Systems (2019, methodology): “[99] adopted data augmentation via sequence preprocessing and embedding dropout to enhance the training process and reduce overfitting, respectively.”
- Deep Learning Based Recommender System (2017, methodology): “Recent advances have demonstrated the efficacy of RNNs in solving this issue [56, 143, 177].”
- Graph Neural Networks in Recommender Systems: A Survey (2020, methodology): “Owing to the advantage of Recurrent Neural Network (RNN) in sequence modeling, some works employ the RNN unit to capture sequence patterns [21, 52].”
- Latent Cross: Making Use of Context in Recurrent Recommender Systems (2018, background): “Researchers have also explored the how preferences evolve in shorter time-scales, e.g., sessions [39].”
- Top-K Off-Policy Correction for a REINFORCE Recommender System (2018, background): “…line of work that is closely re-lated to ours is the growing body of literature on applying deep neural networks to recommender systems [11, 16, 37], in particular using recurrent neural networks to incorporate temporal information and historical events for recommendation [6, 17, 20, 45, 49].”
- A review on deep learning for recommender systems: challenges and remedies (2018, methodology): “As an alternative, session-based recommender systems rely on the recent behavior of users within the current session, which helps to handle cold-start users (Tan et al. 2016).”
- STAMP: Short-Term Attention/Memory Priority Model for Session-based Recommendation (2018, methodology): “Because the Yoochoose training set is quite large and training on the recent fractions yields better results than training on the entire fractions as per the experiments of [17], we use the recent fractions 1/64 and 1/4 of training sequences.”
- ADER: Adaptively Distilled Exemplar Replay Towards Continual Learning for Session-based Recommendation (2020, background): “Based on GRU4Rec, [10] proposes new ranking losses on relevant sessions, and [36] proposes to augment training data.”
- Self-Supervised Hypergraph Convolutional Networks for Session-based Recommendation (2020, background): “(Hidasi et al. 2015; Tan, Xu, and Liu 2016; Li et al. 2017; Liu et al. 2018) are the classical RNNs-based models which borrow the strengths of RNNs to model session-based data.”
- Personalizing Session-based Recommendations with Hierarchical Recurrent Neural Networks (2017, methodology): “In [19] proposed data augmentation techniques to improve the performance of the RNN for session-based recommendations, these techniques have though the side effect of increasing training times as a single session is split into several sub-sessions for training.”
- Recommendation system based on deep learning methods: a systematic review and new directions (2019, methodology): “For example, Tan et al. (2016) exploit RNN architecture for improving session based recommendation by examining and adapting different techniques.”
- Neural Survival Recommender (2017, background): “One work that is closely related to ours was the one in [31] which proposed to solve the ”session-based”recommendation using recurrent networks.”