Optimizing top-n collaborative filtering via dynamic negative item sampling

By Weinan Zhang, Tianqi Chen, Jun Wang, Yong Yu. 2013

306 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Information Systems, Transportation

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

1. Xiangnan He
2. Xiang Wang
3. Fuli Feng
4. Weinan Zhang
5. Julian McAuley
6. Jun Wang
7. Yong Yu
8. Lantao Yu

In Information Systems

1. Xiangnan He
2. Xiang Wang
3. Fuli Feng
4. Julian McAuley
5. Tat-seng Chua
6. Shuai Yuan
7. Weinan Zhang
8. Jun Wang

In Transportation

1. Weinan Zhang
2. Jun Wang
3. Xiangnan He
4. Irwin King
5. Fuli Feng
6. Xiang Wang
7. Chunyan Miao
8. Harald Steck

Citations in the corpus

• Gaussian Ranking by Matrix Factorization (2015, background): “There has been a large body of work on training collaborative filtering approaches as to maximize various ranking metrics, see eg [5, 8, 17, 2, 18, 10, 20, 21, 14, 13, 12].”
• IRGAN: A Minimax Game for Unifying Generative and Discriminative Information Retrieval Models (2017, methodology): “In the experiments, we compare IRGAN with Bayesian Personalised Ranking (BPR) [34] and a state-of-the-art LambdaRank based collaborative €ltering (LambdaFM) [46] for top-N item recommendation tasks [46, 51].”
• Dynamic Attention Deep Model for Article Recommendation by Learning Human Editors' Demonstration (2017, methodology): “CF methods [22, 37, 51] focus on revealing the pa‹erns in the historic user-item interaction data [47] and make recommendations based on the assumption that users with similar behavior like similar items, without considering any a‹ributes or content of the items.”
• Leveraging Social Connections to Improve Personalized Ranking for Collaborative Filtering (2014, methodology): “Here β is a constant and we fix it to 0 as in [24].”
• Bootstrap Latent Representations for Multi-modal Recommendation (2022, result): “For example, previous research [17] has verified that the default uniform sampling [18] in LightGCN [19] obsesses more than 25% training time per epoch.”
• Self-supervised Graph Learning for Recommendation (2020, background): “4, with a proper τ , SGL benefits from dynamic hard negative mining, where the hard negative samples offer meaningful and larger gradients to guide the optimization [54].”
• Optimal real-time bidding for display advertising (2014, methodology): “In this paper, we follow the convention of the offline evaluations from sponsored search [19], recommender systems [39] and Web search [11] that the objects (auctions) with unseen user feedback are ignored (i.e., considered as auction failure cases).”