Logit Averaging: Capturing Global Relation for Session-Based Recommendation
Abstract
:1. Introduction
- By interpolating logits of the sessions with the same target value, we propose a simple yet effective method to incorporate global contexts for session-based recommendation and achieve high accuracy and diversity;
- We conduct extensive experiments on four real-world datasets in six different deep-learning-based session recommendation models and successfully demonstrate the effectiveness of our method;
- We can easily “plug and play” Logit Averaging (LA) in various neural network-based session recommendation models.
2. Proposed Method
2.1. Problem Definition and Notation
2.2. Logit Averaging
Algorithm 1 Logit Averaging. |
Input: sessions S, Short-head Items , Item Encoder , Session Encoder Initialize IE and SE for each iteration do for all do if in target items of sessions then end if end for L is calculated by using Equation (3) end for |
3. Results
3.1. Setup
3.1.1. Datasets
3.1.2. Baseline Methods
3.1.3. Evaluation Metrics
- Recall is the metric to measure recommendation accuracy. It is the proportion of correct prediction amongst top-k recommendation lists.
- Mean Reciprocal Rank (MRR) also measures recommendation accuracy taking ranks into consideration. MRR is the mean of ground truths’ reciprocal rank in the lists.
- Coverage is the ratio of the unique items that appears in recommendation lists.
3.2. Experiment Result
3.3. Ablation Study
3.3.1. Effectiveness of Augmentation
3.3.2. Comparison on GCE-GNN
4. Discussion
Author Contributions
Funding
Institutional Review Board Statement
Informed Consent Statement
Data Availability Statement
Acknowledgments
Conflicts of Interest
Appendix A
Appendix A.1
Dataset | Model | Aug | LA | HR@10 | MRR@10 | Coverage@10 | HR@20 | MRR@20 | Coverage@20 |
---|---|---|---|---|---|---|---|---|---|
RetailRocket | NARM | X | X | 39.24 | 20.76 | 95.52 | 45.92 | 21.24 | 97.63 |
X | O | 39.02 | 21.07 | 95.86 | 45.64 | 21.53 | 97.80 | ||
GraphMix | O | 37.89 | 19.46 | 95.99 | 44.76 | 19.94 | 98.14 | ||
FLAG | O | 39.94 | 21.16 | 95.54 | 46.21 | 21.64 | 97.75 | ||
Random deletion | O | 39.83 | 19.48 | 95.18 | 46.68 | 19.95 | 97.50 | ||
Random Insertion | O | 38.97 | 19.42 | 94.88 | 46.41 | 19.94 | 97.17 | ||
NISER | X | X | 33.24 | 22.24 | 88.66 | 37.54 | 22.53 | 93.42 | |
X | O | 33.33 | 22.09 | 88.59 | 37.66 | 22.39 | 93.39 | ||
GraphMix | O | 33.19 | 22.37 | 85.65 | 37.49 | 22.66 | 90.79 | ||
FLAG | O | 33.40 | 22.45 | 89.56 | 37.76 | 22.75 | 94.40 | ||
Random deletion | O | 34.71 | 17.51 | 96.44 | 39.45 | 17.83 | 98.57 | ||
Random Insertion | O | 32.27 | 16.35 | 96.80 | 36.40 | 16.64 | 98.83 | ||
SR-SAN | X | X | 37.96 | 17.25 | 93.97 | 44.91 | 17.74 | 96.53 | |
X | O | 37.50 | 17.26 | 92.44 | 44.33 | 17.75 | 95.45 | ||
GraphMix | O | 33.89 | 20.13 | 80.22 | 39.14 | 20.49 | 84.66 | ||
FLAG | O | 37.76 | 17.17 | 93.32 | 44.56 | 17.65 | 96.26 | ||
Random deletion | O | 37.42 | 17.01 | 91.90 | 45.57 | 17.58 | 94.86 | ||
Random Insertion | O | 35.06 | 15.97 | 90.53 | 42.61 | 16.48 | 94.10 | ||
TAGNN++ | X | X | 34.30 | 21.54 | 81.59 | 39.83 | 21.92 | 85.47 | |
X | O | 34.10 | 21.71 | 82.23 | 39.61 | 22.09 | 86.21 | ||
GraphMix | O | 38.00 | 23.36 | 90.96 | 43.58 | 23.75 | 93.47 | ||
FLAG | O | 38.35 | 23.18 | 89.47 | 44.47 | 23.61 | 92.62 | ||
Random deletion | O | 36.07 | 19.51 | 82.40 | 42.93 | 19.98 | 86.46 | ||
Random Insertion | O | 31.59 | 16.80 | 77.42 | 38.29 | 17.27 | 82.76 | ||
Tmall | NARM | X | X | 17.14 | 12.52 | 17.68 | 19.18 | 12.71 | 28.68 |
X | O | 16.46 | 12.21 | 17.28 | 18.50 | 12.35 | 28.17 | ||
GraphMix | O | 19.18 | 13.74 | 18.56 | 21.62 | 13.88 | 30.85 | ||
FLAG | O | 16.65 | 12.41 | 17.81 | 18.70 | 12.51 | 28.84 | ||
Random deletion | O | 18.31 | 13.25 | 18.23 | 19.86 | 13.38 | 29.60 | ||
Random Insertion | O | 18.70 | 12.91 | 18.50 | 20.74 | 13.05 | 30.28 | ||
NISER | X | X | 18.89 | 14.04 | 18.93 | 21.03 | 14.18 | 31.83 | |
X | O | 18.11 | 13.52 | 19.21 | 20.35 | 13.60 | 32.43 | ||
GraphMix | O | 19.08 | 13.76 | 18.65 | 20.45 | 13.85 | 31.08 | ||
FLAG | O | 19.18 | 13.80 | 19.03 | 21.13 | 13.93 | 32.01 | ||
Random deletion | O | 18.31 | 13.58 | 18.82 | 19.77 | 13.68 | 31.97 | ||
Random Insertion | O | 16.46 | 12.66 | 18.86 | 18.21 | 12.78 | 31.83 | ||
SR-SAN | X | X | 18.99 | 12.74 | 12.67 | 21.62 | 12.92 | 19.65 | |
X | O | 20.55 | 14.25 | 12.63 | 23.17 | 14.43 | 19.54 | ||
GraphMix | O | 22.40 | 14.65 | 16.27 | 24.83 | 14.83 | 26.07 | ||
FLAG | O | 19.28 | 12.62 | 12.49 | 22.01 | 12.81 | 19.30 | ||
Random deletion | O | 19.28 | 11.84 | 13.12 | 22.59 | 12.03 | 20.45 | ||
Random Insertion | O | 17.92 | 10.55 | 13.00 | 20.84 | 10.76 | 20.12 | ||
TAGNN++ | X | X | 33.11 | 16.17 | 15.88 | 39.82 | 16.64 | 25.01 | |
X | O | 30.77 | 14.55 | 16.36 | 36.42 | 14.96 | 25.42 | ||
GraphMix | O | 34.27 | 17.11 | 17.50 | 43.33 | 17.76 | 28.06 | ||
FLAG | O | 34.86 | 15.38 | 16.56 | 45.47 | 16.14 | 26.01 | ||
Random deletion | O | 30.48 | 14.17 | 16.75 | 36.51 | 14.60 | 26.20 | ||
Random Insertion | O | 21.81 | 10.59 | 15.44 | 26.58 | 10.94 | 23.52 | ||
Yoochoose | NARM | X | X | 59.95 | 29.46 | 29.24 | 70.67 | 30.20 | 35.73 |
X | O | 59.96 | 29.65 | 28.88 | 70.61 | 30.40 | 35.37 | ||
GraphMix | O | 55.69 | 28.08 | 36.91 | 66.66 | 28.85 | 45.52 | ||
FLAG | O | 60.50 | 29.57 | 29.83 | 71.07 | 30.33 | 35.89 | ||
Random deletion | O | 59.99 | 29.38 | 29.17 | 70.54 | 30.13 | 35.75 | ||
Random Insertion | O | 60.49 | 29.68 | 30.50 | 71.00 | 30.42 | 37.09 | ||
NISER | X | X | 60.25 | 30.91 | 29.46 | 70.55 | 31.64 | 34.61 | |
X | O | 60.23 | 30.88 | 29.40 | 70.48 | 31.60 | 34.73 | ||
GraphMix | O | 60.58 | 31.08 | 28.65 | 70.54 | 31.78 | 33.29 | ||
FLAG | O | 60.45 | 30.77 | 29.91 | 70.71 | 31.49 | 35.34 | ||
Random deletion | O | 60.36 | 29.04 | 31.82 | 70.41 | 29.74 | 38.69 | ||
Random Insertion | O | 59.88 | 29.18 | 34.17 | 69.88 | 29.88 | 42.22 | ||
Yoochoose | SR-SAN | X | X | 53.90 | 25.96 | 28.28 | 64.89 | 26.73 | 32.62 |
X | O | 53.75 | 26.00 | 28.31 | 64.91 | 26.77 | 32.63 | ||
GraphMix | O | 54.87 | 26.48 | 30.19 | 66.14 | 27.27 | 34.88 | ||
FLAG | O | 54.31 | 26.10 | 28.40 | 65.24 | 26.86 | 32.86 | ||
Random deletion | O | 55.54 | 25.85 | 28.53 | 67.21 | 26.65 | 32.61 | ||
Random Insertion | O | 52.28 | 23.36 | 28.99 | 65.54 | 24.27 | 33.08 | ||
TAGNN++ | X | X | 60.70 | 30.63 | 26.48 | 70.93 | 31.35 | 30.72 | |
X | O | 60.84 | 31.11 | 28.79 | 71.16 | 31.83 | 33.46 | ||
GraphMix | O | 60.75 | 30.86 | 27.64 | 71.15 | 31.59 | 32.27 | ||
FLAG | O | 60.67 | 30.95 | 27.82 | 71.08 | 31.68 | 31.72 | ||
Random deletion | O | 58.32 | 29.43 | 16.79 | 68.79 | 29.17 | 18.96 | ||
Random Insertion | O | 57.32 | 27.88 | 17.03 | 68.44 | 28.65 | 19.12 | ||
Diginetica | NARM | X | X | 35.86 | 18.60 | 68.78 | 44.83 | 19.22 | 82.60 |
X | O | 35.76 | 18.51 | 69.05 | 44.49 | 19.12 | 83.07 | ||
GraphMix | O | 34.76 | 18.38 | 69.02 | 43.39 | 18.97 | 83.16 | ||
FLAG | O | 36.30 | 18.49 | 69.06 | 45.40 | 19.12 | 82.91 | ||
Random deletion | O | 35.96 | 18.61 | 68.95 | 45.32 | 19.25 | 82.66 | ||
Random Insertion | O | 36.59 | 18.71 | 68.31 | 45.59 | 19.36 | 82.13 | ||
NISER | X | X | 34.12 | 17.98 | 65.02 | 41.89 | 18.52 | 77.34 | |
X | O | 33.79 | 18.02 | 65.28 | 41.68 | 18.57 | 78.25 | ||
GraphMix | O | 34.27 | 18.06 | 59.23 | 42.04 | 18.59 | 70.59 | ||
FLAG | O | 34.04 | 18.10 | 65.41 | 41.69 | 18.63 | 78.41 | ||
Random deletion | O | 33.55 | 18.37 | 64.18 | 39.95 | 18.81 | 77.65 | ||
Random Insertion | O | 33.11 | 18.28 | 64.04 | 39.68 | 18.74 | 77.69 | ||
SR-SAN | X | X | 32.54 | 16.61 | 67.60 | 41.50 | 17.22 | 78.56 | |
X | O | 32.93 | 16.93 | 66.68 | 42.02 | 17.56 | 77.34 | ||
GraphMix | O | 32.94 | 16.95 | 70.66 | 41.72 | 17.56 | 81.55 | ||
FLAG | O | 32.62 | 16.62 | 67.57 | 41.28 | 17.22 | 78.39 | ||
Random deletion | O | 32.07 | 14.85 | 67.21 | 42.44 | 15.58 | 79.28 | ||
Random Insertion | O | 31.50 | 14.04 | 66.78 | 41.87 | 14.76 | 79.36 | ||
TAGNN++ | X | X | 33.90 | 17.34 | 62.81 | 42.64 | 17.95 | 74.00 | |
X | O | 36.42 | 18.31 | 68.85 | 45.41 | 18.93 | 79.83 | ||
GraphMix | O | 36.36 | 18.43 | 70.89 | 45.02 | 19.03 | 80.91 | ||
FLAG | O | 33.77 | 17.36 | 62.54 | 42.29 | 17.96 | 73.66 | ||
Random deletion | O | 34.17 | 15.37 | 66.93 | 43.61 | 16.03 | 78.87 | ||
Random Insertion | O | 34.20 | 18.02 | 63.99 | 41.97 | 18.56 | 76.07 |
Appendix A.2
Hyperparameter | Model | |||||
---|---|---|---|---|---|---|
NARM | EOPA | NISER | SRGNN | SRSAN | TAGNN++ | |
Batch Size | 128 | 128 | 128 | 128 | 128 | 64 |
Hidden Size | 100 | - | 100 | 100 | 96 | 100 |
Epoch | 100 | 30 | 30 | 30 | 30 | 30 |
L2 Penalty | - | 0.0001 | 0.00001 | 0.00001 | 0.00001 | 0.00001 |
Learning rate | 0.001 | 0.001 | 0.001 | 0.001 | 0.001 | 0.001 |
Learning rate decay rate | 0.1 | - | 0.1 | 0.1 | 0.1 | 0.1 |
Number of steps after which the learning rate decay | 80 | - | 3 | 3 | 3 | 3 |
The dropout ratio for features | - | 0.2 | - | - | - | - |
GNN Propagation | - | - | 1 | 1 | 1 | 1 |
Number of SAN layer | - | - | - | - | 1 | - |
Number of heads of multi-head attention | - | - | - | - | 2 | - |
Multipler of hidden size | - | - | - | - | 1 | - |
Embedding Dimension | 50 | 32 | - | - | - | |
Number of layers | 1 | 3 | - | - | - | - |
Appendix A.3
References
- Zhang, S.; Yao, L.; Sun, A.; Tay, Y. Deep learning based recommender system: A survey and new perspectives. arXiv 2019, arXiv:1707.07435. [Google Scholar] [CrossRef] [Green Version]
- Covington, P.; Adams, J.; Sargin, E. Deep neural networks for youtube recommendations. In Proceedings of the 10th ACM Conference on Recommender Systems, Boston, MA, USA, 15–19 September 2016; pp. 191–198. [Google Scholar]
- Sarwar, B.; Karypis, G.; Konstan, J.; Riedl, J. Item-Based Collaborative Filtering Recommendation Algorithms. In Proceedings of the 10th International Conference on World Wide Web, Hong Kong, China, 1–5 May 2001; pp. 285–295. [Google Scholar]
- Rendle, S.; Freudenthaler, C.; Schmidt-Thieme, L. Factorizing Personalized Markov Chains for Next-Basket Recommendation. In Proceedings of the 19th International Conference on World Wide Web, Raleigh, NC, USA, 26–30 April 2010; pp. 811–820. [Google Scholar]
- He, X.; Liao, L.; Zhang, H.; Nie, L.; Hu, X.; Chua, T. Neural Collaborative Filtering. In Proceedings of the 26th International Conference on World Wide Web, Perth, Australia, 3–7 April 2017; pp. 173–182. [Google Scholar]
- Chen, Y.; de Rijke, M. A collective variational autoencoder for top-n recommendation with side information. In Proceedings of the 3rd Workshop on Deep Learning for Recommender Systems, Vancouver, BC, Canada, 6 October 2018; pp. 3–9. [Google Scholar]
- Dong, X.; Yu, L.; Wu, Z.; Sun, Y.; Yuan, L.; Zhang, F. A hybrid collaborative filtering model with deep structure for recommender systems. In Proceedings of the AAAI Conference on Artificial Intelligence, San Francisco, CA, USA, 4–9 February 2017. [Google Scholar]
- Liu, F.; Cheng, Z.; Zhu, L.; Gao, Z.; Nie, L. Interest-aware message-passing gcn for recommendation. In Proceedings of the Web Conference 2021, Ljubljana, Solvenia, 19–23 April 2021; pp. 1296–1305. [Google Scholar]
- Choi, M.; Kim, J.; Lee, J.; Shim, H.; Lee, J. Session-Aware Linear Item-Item Models for Session-Based Recommendation. In Proceedings of the Web Conference 2021, Ljubljana, Slovenia, 19–23 April 2021; pp. 2186–2197. [Google Scholar]
- Wang, S.; Cao, L.; Wang, Y.; Sheng, Q.Z.; Orgun, M.A.; Lian, D. A Survey on Session-Based Recommender Systems. arXiv 2021, arXiv:1902.04864. [Google Scholar] [CrossRef]
- Li, J.; Ren, P.; Chen, Z.; Ren, Z.; Lian, T.; Ma, J. Neural attentive session-based recommendation. In Proceedings of the 2017 ACM on Conference on Information and Knowledge Management, Singapore, 6–10 November 2017; pp. 1419–1428. [Google Scholar]
- Fang, J. Session-based Recommendation with Self-Attention Networks. arXiv 2021, arXiv:2102.01922. [Google Scholar]
- Cho, K.; Merrienboer, V.B.; Gulcehre, C.; Bahdanau, D.; Bougares, F.; Schwenk, H.; Bengio, Y. Learning Phrase Representations using RNN Encoder–Decoder for Statistical Machine Translation. In Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing, Doha, Qatar, 25–29 October 2014; pp. 1724–1734. [Google Scholar]
- Kipf, T.N.; Welling, M. Semi-Supervised Classification with Graph Convolutional Networks. In Proceedings of the 5th International Conference on Learning Representations, Toulon, France, 24–26 April 2017. [Google Scholar]
- Wu, S.; Tang, Y.; Zhu, Y.; Wang, L.; Xie, X.; Tan, T. Session-based recommendation with graph neural networks. In Proceedings of the Thirty-Third AAAI Conference on Artificial Intelligence, Honolulu, HI, USA, 27 January–1 February 2019. [Google Scholar]
- Gupta, P.; Garg, D.; Malhotra, P.; Vig, L.; Shroff, G. NISER: Normalized item and session representations to handle popularity bias. In Proceedings of the 1st International Workshop on Graph Representation Learning on Its Applications (CIKM ’19), Beijing, China, 3–7 November 2019. [Google Scholar]
- Chen, T.; Wong, R.C.W. Handling information loss of graph neural networks for session-based recommendation. In Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, Virtual Event, 23–27 August 2020; pp. 1172–1180. [Google Scholar]
- Mitheran, S.; Java, A.; Sahu, S.K.; Shaikh, A. Introducing Self-Attention to Target Attentive Graph Neural Networks. arXiv 2022, arXiv:2107.01516. [Google Scholar]
- Li, A.; Cheng, Z.; Liu, F.; Gao, Z.; Guan, W.; Peng, Y. Disentangled Graph Neural Networks for Session-based Recommendation. arXiv 2022, arXiv:2201.03482. [Google Scholar]
- Li, Y.; Tarlow, D.; Brockschmidt, M.; Zemel, R.S. Gated Graph Sequence Neural Networks. In Proceedings of the 4th International Conference on Learning Representations, San Juan, Puerto Rico, 2–4 May 2016. [Google Scholar]
- Wang, M.; Ren, P.; Mei, L.; Chen, Z.; Ma, J.; de Rijke, M. A Collaborative Session-Based Recommendation Approach with Parallel Memory Modules. In Proceedings of the 42nd International ACM SIGIR Conference on Research and Development in Information Retrieval, Paris, France, 21–25 July 2019; pp. 345–354.
- Wang, Z.; Wei, W.; Cong, G.; Li, X.L.; Mao, X.L.; Qiu, M. Global context enhanced graph neural networks for session-based recommendation. In Proceedings of the 43rd International ACM SIGIR Conference on Research and Development in Information Retrieval, Virtual Event, 25–30 July 2020; pp. 169–178. [Google Scholar]
- Mi, F.; Faltings, B. Memory Augmented Neural Model for Incremental Session-Based Recommendation. In Proceedings of the Twenty-Ninth International Joint Conference on Artificial Intelligence, Yokohama, Japan, 7–15 January 2020; pp. 2169–2176. [Google Scholar]
- Zhou, H.; Tan, Q.; Huang, X.; Zhou, K.; Wang, X. Temporal Augmented Graph Neural Networks for Session-Based Recommendations. In Proceedings of the 44th International ACM SIGIR Conference on Research and Development in Information Retrieval, Virtual Event, 11–15 July 2021; pp. 1798–1802. [Google Scholar]
- Huang, C.; Xia, X.; Dai, C.; Bo, Z. Graph-Enhanced Multi-Task Learning of Multi-Level Transition Dynamics for Session-based Recommendation. In Proceedings of the 35th AAAI Conference on Artificial Intelligence, Virtual Event, 2–9 February 2021; pp. 4123–4130. [Google Scholar]
- Xia, X.; Yin, H.; Yu, J.; Shao, Y.; Cui, L. Self-Supervised Graph Co-Training for Session-based Recommendation. Assoc. Comput. Mach. 2021, 11, 2180–2190. [Google Scholar]
- Pang, Y.; Wu, L.; Shen, Q.; Zhang, Y.; Wei, Z.; Xu, F.; Chang, E.; Long, B.; Pei, J. Heterogeneous global graph neural networks for personalized session-based recommendation. In Proceedings of the Fifteenth ACM International Conference on Web Search and Data Mining, Virtual Event, 21–25 February 2022; pp. 775–783. [Google Scholar]
- Menon, A.; Jayasumana, S.; Rawat, A.; Jain, H.; Veit, A.; Kumar, S. Long-tail learning via logit adjustment. In Proceedings of the 2021 International Conference on Learning Representation, Virtual Conference, 3–7 May 2021. [Google Scholar]
- Hurley, N.; Zhang, M. Novelty and Diversity in Top-N Recommendation—Analysis and Evaluation. ACM Trans. Internet Technol. 2011, 10, 14. [Google Scholar] [CrossRef]
- Bradley, K.; Smyth, B. Improving recommendation diversity. In Proceedings of the Twelfth Irish Conference on Artificial Intelligence and Cognitive Science, Maynooth, Ireland, 1 September 2001; pp. 141–152. [Google Scholar]
- Miyamoto, S.; Zamami, T.; Yamana, H. Appearance frequency-based ranking method for improving recommendation diversity. In Proceedings of the 2019 IEEE 4th International Conference on Big Data Analytics (ICBDA), Suzhou, China, 15–18 March 2019; pp. 420–425. [Google Scholar]
- Wang, L.; Zhang, X.; Wang, T.; Wan, S.; Srivastava, G.; Pang, S.; Qi, L. Diversified and scalable service recommendation with accuracy guarantee. IEEE Trans. Comput. Soc. Syst. 2020, 8, 1182–1193. [Google Scholar] [CrossRef]
- Anderson, C. The Long Tail: Why the Future of Business is Selling Less of More; Hachette Books: London, UK, 2006. [Google Scholar]
- Park, Y.J.; Tuzhilin, A. The long tail of recommender systems and how to leverage it. In Proceedings of the 2008 ACM Conference on Recommender Systems, Lausanne, Switzerland, 23–25 October 2008; pp. 11–18. [Google Scholar]
- Isufi, E.; Pocchiari, M.; Hanjalic, A.T. Accuracy-diversity trade-off in recommender systems via graph convolutions. Inf. Process. Manag. 2021, 58, 102459. [Google Scholar] [CrossRef]
- Wang, Y.; Wang, W.; Liang, Y.; Cai, Y.; Hooi, B. Mixup for Node and Graph Classification. In Proceedings of the Web Conference 2021, Ljubljana, Slovenia, 12–16 April 2021; pp. 3663–3674. [Google Scholar]
- Kong, K.; Li, G.; Ding, M.; Wu, Z.; Zhu, C.; Ghanem, B.; Taylor, G.; Goldstein, T. Flag: Adversarial data augmentation for graph neural networks. arXiv 2020, arXiv:2010.09891. [Google Scholar]
Statistics | Diginetica | Retailrocket | Yoochoose | Tmall |
---|---|---|---|---|
# Train Sessions | 186,670 | 113,287 | 124,472 | 65,286 |
# Test Sessions | 15,936 | 71,235 | 15,324 | 1027 |
# Items | 43,098 | 27,413 | 37,484 | 40,728 |
Avg. Length | 4.85 | 3.66 | 3.97 | 6.68 |
Dataset | Model | LA | HR@10 | MRR@10 | Coverage@10 | HR@20 | MRR@20 | Coverage@20 |
---|---|---|---|---|---|---|---|---|
RetailRocket | NARM | X | 39.24 | 20.76 | 95.52 | 45.92 | 21.24 | 97.63 |
EOPA | 33.74 | 17.13 | 81.07 | 42.51 | 17.72 | 86.91 | ||
NISER | 33.24 | 22.24 | 88.66 | 37.54 | 22.53 | 93.42 | ||
SR-GNN | 31.98 | 18.81 | 76.09 | 37.24 | 19.17 | 80.40 | ||
SR-SAN | 37.96 | 17.25 | 93.97 | 44.91 | 17.74 | 96.53 | ||
TAGNN++ | 34.30 | 21.54 | 81.59 | 39.83 | 21.92 | 85.47 | ||
NARM | O | 39.02 | 21.07 | 95.86 | 45.64 | 21.53 | 97.80 | |
EOPA | 33.63 | 17.33 | 81.28 | 42.28 | 17.91 | 86.98 | ||
NISER | 33.33 | 22.09 | 88.59 | 37.66 | 22.39 | 93.39 | ||
SR-GNN | 34.40 | 21.92 | 85.34 | 39.45 | 22.27 | 89.29 | ||
SR-SAN | 37.50 | 17.26 | 92.44 | 44.33 | 17.75 | 95.45 | ||
TAGNN++ | 34.10 | 21.71 | 82.23 | 39.61 | 22.09 | 86.21 | ||
Gain (%) | 0.86 | 3.23 | 1.97 | 0.54 | 3.14 | 1.84 | ||
Tmall | NARM | X | 17.14 | 12.52 | 17.68 | 19.18 | 12.71 | 28.68 |
EOPA | 19.58 | 9.70 | 32.27 | 24.64 | 10.02 | 44.43 | ||
NISER | 18.89 | 14.04 | 18.93 | 21.03 | 14.18 | 31.83 | ||
SR-GNN | 16.55 | 11.00 | 13.79 | 18.79 | 11.17 | 21.52 | ||
SR-SAN | 18.99 | 12.74 | 12.67 | 21.62 | 12.92 | 19.65 | ||
TAGNN++ | 33.11 | 16.17 | 15.88 | 39.82 | 16.64 | 25.01 | ||
NARM | O | 16.46 | 12.21 | 17.28 | 18.5 | 12.35 | 28.17 | |
EOPA | 19.43 | 9.81 | 31.62 | 24.14 | 10.14 | 43.66 | ||
NISER | 18.11 | 13.52 | 19.21 | 20.35 | 13.60 | 32.43 | ||
SR-GNN | 16.85 | 11.92 | 15.47 | 19.67 | 12.11 | 24.81 | ||
SR-SAN | 20.55 | 14.25 | 12.63 | 23.17 | 14.43 | 19.54 | ||
TAGNN++ | 30.77 | 14.55 | 16.36 | 36.42 | 14.96 | 25.42 | ||
Gain (%) | −0.98 | 0.86 | 2.02 | −0.92 | 0.71 | 2.46 | ||
Yoochoose | NARM | X | 59.95 | 29.46 | 29.24 | 70.67 | 30.20 | 35.73 |
EOPA | 51.83 | 24.91 | 28.81 | 62.85 | 25.68 | 31.74 | ||
NISER | 60.25 | 30.91 | 29.46 | 70.55 | 31.64 | 34.61 | ||
SR-GNN | 59.56 | 29.47 | 23.19 | 69.87 | 30.19 | 26.54 | ||
SR-SAN | 53.90 | 25.96 | 28.28 | 64.89 | 26.73 | 32.62 | ||
TAGNN++ | 60.70 | 30.63 | 26.48 | 70.93 | 31.35 | 30.72 | ||
NARM | O | 59.96 | 29.65 | 28.88 | 70.61 | 30.40 | 35.37 | |
EOPA | 51.54 | 24.90 | 29.02 | 62.64 | 25.68 | 31.73 | ||
NISER | 60.23 | 30.88 | 29.40 | 70.48 | 31.60 | 34.73 | ||
SR-GNN | 60.46 | 30.90 | 27.68 | 70.55 | 31.61 | 32.06 | ||
SR-SAN | 53.75 | 26.00 | 28.31 | 64.91 | 26.77 | 32.63 | ||
TAGNN++ | 60.84 | 31.11 | 28.79 | 71.16 | 31.83 | 33.46 | ||
Gain (%) | 0.15 | 1.18 | 4.58 | 0.14 | 1.15 | 4.84 | ||
Diginetica | NARM | X | 35.86 | 18.60 | 68.78 | 44.83 | 19.22 | 82.60 |
EOPA | 34.25 | 14.52 | 62.09 | 47.09 | 15.37 | 75.91 | ||
NISER | 34.12 | 17.98 | 65.02 | 41.89 | 18.52 | 77.34 | ||
SR-GNN | 32.84 | 16.83 | 60.95 | 41.46 | 17.43 | 71.44 | ||
SR-SAN | 32.54 | 16.61 | 67.60 | 41.50 | 17.22 | 78.56 | ||
TAGNN++ | 33.90 | 17.34 | 62.81 | 42.64 | 17.95 | 74.00 | ||
NARM | O | 35.76 | 18.51 | 69.05 | 44.49 | 19.12 | 83.07 | |
EOPA | 34.54 | 14.60 | 62.33 | 47.03 | 15.47 | 75.81 | ||
NISER | 33.79 | 18.02 | 65.28 | 41.68 | 18.57 | 78.25 | ||
SR-GNN | 33.85 | 17.72 | 64.47 | 42.03 | 18.29 | 76.34 | ||
SR-SAN | 32.93 | 16.93 | 66.68 | 42.02 | 17.56 | 77.34 | ||
TAGNN++ | 36.42 | 18.31 | 68.85 | 45.41 | 18.93 | 79.83 | ||
Gain (%) | 1.88 | 2.18 | 2.53 | 1.29 | 2.13 | 2.47 | ||
Total Gain (%) | 0.48 | 1.86 | 2.78 | 0.26 | 1.78 | 2.90 |
Dataset | Augmentation | LA | HR@10 | MRR@10 | Coverage@10 | HR@20 | MRR@20 | Coverage@20 |
---|---|---|---|---|---|---|---|---|
RetailRocket | X | O | −0.20 | 0.09 | −0.15 | −0.24 | 0.08 | −0.05 |
GraphMix | O | −0.44 | 0.88 | −1.73 | −0.81 | 0.85 | −1.50 | |
FLAG | O | 1.18 | 0.54 | 2.04 | 1.20 | 0.56 | 2.00 | |
Random deletion | O | 0.82 | −2.07 | 1.55 | 1.61 | −2.02 | 1.09 | |
Random Insertion | O | −1.71 | −3.31 | −0.03 | −1.12 | −3.28 | −0.05 | |
Tmall | X | O | −0.56 | −0.24 | 0.08 | −0.80 | −0.28 | 0.10 |
GraphMix | O | 1.70 | 0.95 | 1.46 | 2.15 | 0.97 | 2.72 | |
FLAG | O | 0.46 | −0.32 | 0.18 | 1.42 | −0.27 | 0.25 | |
Random deletion | O | −0.44 | −0.66 | 0.44 | −0.73 | −0.69 | 0.76 | |
Random Insertion | O | −3.31 | −2.19 | 0.16 | −3.82 | −2.23 | 0.15 | |
Yoochoose | X | O | −0.01 | 0.17 | 0.48 | 0.03 | 0.17 | 0.63 |
GraphMix | O | −0.73 | −0.12 | 2.48 | −0.64 | −0.11 | 3.07 | |
FLAG | O | 0.28 | 0.11 | 0.62 | 0.26 | 0.11 | 0.53 | |
Random deletion | O | −0.15 | −0.82 | −1.79 | −0.02 | −1.06 | −1.92 | |
Random Insertion | O | −1.21 | −1.72 | −0.69 | −0.55 | −1.68 | −0.54 | |
Diginetica | X | O | 0.62 | 0.31 | 1.41 | 0.69 | 0.32 | 1.50 |
GraphMix | O | 0.48 | 0.32 | 1.40 | 0.33 | 0.31 | 0.93 | |
FLAG | O | 0.08 | 0.01 | 0.09 | −0.05 | 0.01 | 0.22 | |
Random deletion | O | −0.17 | −0.83 | 0.77 | 0.12 | −0.81 | 1.49 | |
Random Insertion | O | −0.25 | −0.37 | −0.27 | −0.44 | −0.37 | 0.69 |
Dataset | Model | HR@10 | MRR@10 | Coverage@10 | HR@20 | MRR@20 | Coverage@20 |
---|---|---|---|---|---|---|---|
Tmall | GCE-GNN | 23.76 | 15.20 | 16.18 | 26.29 | 15.35 | 25.71 |
TAGNN++ | 30.77 | 14.55 | 16.36 | 36.42 | 14.96 | 25.42 | |
TAGNN++ w. GraphMix | 34.27 | 17.11 | 17.50 | 43.33 | 17.76 | 28.06 | |
Yoochoose | GCE-GNN | 60.32 | 29.66 | 29.79 | 71.02 | 30.40 | 34.98 |
TAGNN++ | 60.84 | 31.11 | 28.79 | 71.16 | 31.83 | 33.46 | |
TAGNN++ w. GraphMix | 60.75 | 30.86 | 27.64 | 71.15 | 31.59 | 32.27 |
Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations. |
© 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).
Share and Cite
Yang, H.; Kim, G.; Lee, J.-H. Logit Averaging: Capturing Global Relation for Session-Based Recommendation. Appl. Sci. 2022, 12, 4256. https://doi.org/10.3390/app12094256
Yang H, Kim G, Lee J-H. Logit Averaging: Capturing Global Relation for Session-Based Recommendation. Applied Sciences. 2022; 12(9):4256. https://doi.org/10.3390/app12094256
Chicago/Turabian StyleYang, Heeyoon, Gahyung Kim, and Jee-Hyoung Lee. 2022. "Logit Averaging: Capturing Global Relation for Session-Based Recommendation" Applied Sciences 12, no. 9: 4256. https://doi.org/10.3390/app12094256
APA StyleYang, H., Kim, G., & Lee, J.-H. (2022). Logit Averaging: Capturing Global Relation for Session-Based Recommendation. Applied Sciences, 12(9), 4256. https://doi.org/10.3390/app12094256