Recurrent Neural Networks with Top-k Gains for Session-based Recommendations
By Balázs Hidasi, Alexandros Karatzoglou. 2017
978 citations recorded by the source corpus.
RNNs have been shown to be excellent models for sequential data and in particular for data that is generated by users in an session-based manner. The use of RNNs provides impressive performance benefits over classical methods in session-based recommendations. In this work we introduce novel ranking loss functions tailored to RNNs in the recommendation setting. The improved performance of these losses over alternatives, along with further tricks and refinements described in this work, allow for an overall improvement of up to 35% in terms of MRR and Recall@20 over previous session-based RNN solutions and up to 53% over classical collaborative filtering approaches. Unlike data augmentation-based improvements, our method does not increase training times significantly. We further demonstrate the performance gain of the RNN over baselines in an online A/B test.
Subfields: Artificial Intelligence, Information Systems, Management Science and Operations Research, 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
In Information Systems
In Management Science and Operations Research
- Dietmar Jannach
- Malte Ludewig
- Xiangnan He
- Julian McAuley
- Wang-Cheng Kang
- Longbing Cao
- Yan Wang
- Quan Z. Sheng
In Model Architectures
- Julian McAuley
- Dietmar Jannach
- Malte Ludewig
- Wang-Cheng Kang
- Hongzhi Yin
- Jiacheng Li
- Yujie Wang
- Xiangnan He
In Recommender Systems
- Xiangnan He
- Julian McAuley
- Wang-Cheng Kang
- Yi Tay
- Dietmar Jannach
- Malte Ludewig
- Hongzhi Yin
- Jiacheng Li
Citations in the corpus
- Variational Autoencoders for Collaborative Filtering (2018, methodology): “dation [8, 15, 16, 42, 45] and in feedward neural networks applied to Youtube recommendation [9].”
- A Simple Convolutional Generative Network for Next Item Recommendation (2018, methodology): “We evaluate the prediction accuracy of the last (i.e., next) item of each sequence in the testing set, similarly to [14].”
- AMER: Automatic Behavior Modeling and Interaction Exploration in Recommender System (2020, methodology): “• GRU4Rec+ [26] is an improved version of GRU4Rec, which utilizes delicately-designed loss function and sampling strategy to train GRU model.”
- Hierarchical Gating Networks for Sequential Recommendation (2019, methodology): “s. For instance, Tang et al. [34] propose to apply the convolutional neural network (CNN) on item embedding sequence, where the short-term contexts can be captured by the convolutional operations. In [12, 13, 20, 28], recurrent neural network (RNN), especially gated recurrent unit (GRU), based metho”
- Self-Supervised Hypergraph Convolutional Networks for Session-based Recommendation (2020, background): “…the research efforts in this area regard the sessions as ordered sequences, among which recurrent neural networks (RNNs) based (Hidasi et al. 2015; Jannach and Ludewig 2017; Hidasi and Karatzoglou 2018) and graph neural networks (GNNs) (Wu et al. 2020) based approaches have shown great performance.”
- Graph Neural Networks in Recommender Systems: A Survey (2020, background): “When the timestamps of the user’s historical behavior are known or the historical behavior is organized in chronological order, the user representations can be enhanced via exploring the sequential patterns in her/his historical interactions [53, 61, 70, 85, 97, 119, 131, 136, 150].”
- Deep Learning Based Recommender System (2017, background): “RNNs [5, 28, 35, 56, 57, 73, 78, 90, 118, 133, 140, 143, 175–177], [24, 29, 33, 55, 68, 91, 108, 113, 134, 142, 150, 174, 180] RBM [42, 71, 72, 101, 124, 168, 181] NADE [36, 204, 205]”
- Rethinking the Item Order in Session-based Recommendation with Graph Neural Networks (2019, methodology): “Based on the RNN model, some work makes improvements on the architectural choice and the objective function design [8, 30].”
- Self-Attentive Sequential Recommendation (2018, methodology): “• GRU4Rec+ [26]: An improved version of GRU4Rec, which adopts a different loss function and sampling strategy, and shows significant performance gains on TopN recommendation.”
- A Survey on Session-based Recommender Systems (2019, methodology): “This definition covers those narrowly conceived SBRSs in some studies [37, 84] that recommend the next interaction in the current session only.”
- SSE-PT: Sequential Recommendation Via Personalized Transformer (2020, methodology): “• GRU4Rec + : follow-up work of GRU4Rec by the same authors: the model has a very similar architecture to GRU4Rec but has a more complicated loss function [11].”
- A historical perspective of explainable Artificial Intelligence (2020, methodology): “…deep learning techniques, such as CNNs; Seo, Huang, Yang, & Liu, 2017) or Recurrent Neural Networks and Long-Short Term Memory networks (RNN-LSTM) (Hidasi & Karatzoglou, 2018) are used to implement different recommendation strategies, such as sequential recommendations (LSTMs) or context-aware…”
- Time Interval Aware Self-Attention for Sequential Recommendation (2020, methodology): “Recurrent Neural Networks (RNNs) [11, 12] have also been applied in this setting.”
- ADER: Adaptively Distilled Exemplar Replay Towards Continual Learning for Session-based Recommendation (2020, background): “Various recurrent neural networks [10, 11] and attention mechanisms [16, 18, 20] have been proposed for φ, and the common choices for the output layer ω is fully-connect layers[11] or bi-linear decoders [16, 18].”
- 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].”
- Evaluation of session-based recommendation algorithms (2018, methodology): “Except for the RSC15-S dataset, which we include to make our evaluation comparable with previous works [Hidasi and Karatzoglou 2017; Jannach and Ludewig 2017], we report the average values after creating five data splits as described above.”
- Next-item Recommendation with Sequential Hypergraphs (2020, methodology): “NDCG For HPMN, GRU4Rec+, HGN and BERT4Rec, we use the implementations and settings as provided in the original papers.”
- Deep Interest Evolution Network for Click-Through Rate Prediction (2018, methodology): “GRU overcomes the vanishing gradients problem of RNN and is faster than LSTM (Hochreiter and Schmidhuber 1997), which is suitable for e-commerce system.”