pith. sign in

arxiv: 1906.11171 · v1 · pith:JHOJQ6DYnew · submitted 2019-06-26 · 💻 cs.IR · cs.LG

Modeling Embedding Dimension Correlations via Convolutional Neural Collaborative Filtering

Xiaoyu Du , Xiangnan He , Fajie Yuan , Jinhui Tang , Zhiguang Qin , Tat-Seng Chua This is my paper

Pith reviewed 2026-05-25 15:02 UTC · model grok-4.3

classification 💻 cs.IR cs.LG
keywords collaborative filteringneural networksembedding correlationsouter productconvolutional neural networkrecommender systemsuser-item interactions
0
0 comments X

The pith

ConvNCF models pairwise and high-order correlations among embedding dimensions by applying outer product to user and item embeddings followed by a convolutional network.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper sets out to show that neural collaborative filtering becomes more effective when correlations among embedding dimensions are modeled explicitly rather than left implicit. It does this by first computing the outer product of a user's embedding and an item's embedding, which surfaces all pairwise dimension interactions, and then feeding the resulting matrix into a convolutional neural network that extracts higher-order patterns. Three versions of the model are tested on two real-world datasets and shown to outperform several competitive neural and non-neural baselines. A sympathetic reader would care because current neural recommenders typically treat embedding dimensions as independent, which may limit how accurately they predict preferences from sparse user-item history.

Core claim

ConvNCF is a neural collaborative filtering framework that applies the outer product operation to user and item embeddings to explicitly model pairwise correlations between embedding dimensions and then employs a convolutional neural network to learn high-order correlations among those dimensions, resulting in improved modeling of user-item affinities.

What carries the argument

Outer product of user and item embeddings followed by a convolutional neural network to extract correlations.

If this is right

  • The model outperforms several competitive CF methods on two real-world datasets.
  • Modeling embedding dimension correlations improves effectiveness in collaborative filtering.
  • Three different instantiations using varied user inputs all benefit from the outer-product-plus-CNN design.
  • The framework provides a general way to capture dimension-wise interactions in neural recommender models.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same outer-product-plus-CNN pattern could be tested in other embedding-based tasks such as knowledge-graph link prediction.
  • It raises whether MLP-based interaction functions in neural CF are limited precisely because they do not treat dimensions as having explicit pairwise structure.
  • Performance differences might vary with interaction sparsity or with the way embeddings are initialized.

Load-bearing premise

That explicitly modeling pairwise and high-order correlations among embedding dimensions via outer product and CNN will produce meaningfully better user-item affinity predictions than existing neural CF architectures on real-world data.

What would settle it

An experiment in which a standard neural CF baseline without the outer-product or CNN layers matches or exceeds ConvNCF accuracy on the same two datasets while controlling for embedding size and training procedure would falsify the claim.

Figures

Figures reproduced from arXiv: 1906.11171 by Fajie Yuan, Jinhui Tang, Tat-Seng Chua, Xiangnan He, Xiaoyu Du, Zhiguang Qin.

Figure 1
Figure 1. Figure 1: An illustration of our proposed Convolutional Neural Collaborative Filtering (ConvNCF) solution. Following the embedding layer is an outer product layer, which generates a 2D matrix (interaction map) that explicitly captures the pairwise correlations between embedding dimensions. The interaction map is then fed into a CNN to model high-order correlations to obtain the final prediction. 1 INTRODUCTION Recom… view at source ↗
Figure 2
Figure 2. Figure 2: Illustration of the embedding function of the three ConvNCF methods. [PITH_FULL_IMAGE:figures/full_fig_p013_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: HR@10 and NDCG@10 of ConvNCF models and corresponding embedding models ( [PITH_FULL_IMAGE:figures/full_fig_p017_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: HR@10 and NDCG@10 of applying different operations above the embedding layer in each epoch [PITH_FULL_IMAGE:figures/full_fig_p018_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: HR@10 and NDCG@10 of using different hidden layers above the interaction map (ConvNCF uses a [PITH_FULL_IMAGE:figures/full_fig_p019_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Performance of ConvNCF-MF w.r.t. different numbers of feature maps per convolutional layer (denoted [PITH_FULL_IMAGE:figures/full_fig_p019_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: HR@10 and NDCG@10 on Yelp via different training tricks: training from scratch and training with [PITH_FULL_IMAGE:figures/full_fig_p020_7.png] view at source ↗
read the original abstract

As the core of recommender system, collaborative filtering (CF) models the affinity between a user and an item from historical user-item interactions, such as clicks, purchases, and so on. Benefited from the strong representation power, neural networks have recently revolutionized the recommendation research, setting up a new standard for CF. However, existing neural recommender models do not explicitly consider the correlations among embedding dimensions, making them less effective in modeling the interaction function between users and items. In this work, we emphasize on modeling the correlations among embedding dimensions in neural networks to pursue higher effectiveness for CF. We propose a novel and general neural collaborative filtering framework, namely ConvNCF, which is featured with two designs: 1) applying outer product on user embedding and item embedding to explicitly model the pairwise correlations between embedding dimensions, and 2) employing convolutional neural network above the outer product to learn the high-order correlations among embedding dimensions. To justify our proposal, we present three instantiations of ConvNCF by using different inputs to represent a user and conduct experiments on two real-world datasets. Extensive results verify the utility of modeling embedding dimension correlations with ConvNCF, which outperforms several competitive CF methods.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The manuscript proposes ConvNCF, a neural collaborative filtering framework that applies an outer product to user and item embeddings to explicitly capture pairwise correlations among embedding dimensions and stacks a CNN on the resulting interaction map to learn high-order correlations. Three instantiations are presented (differing in user representation) and evaluated on two real-world datasets, where ConvNCF is reported to outperform several competitive neural and non-neural CF baselines.

Significance. If the performance gains are shown to arise specifically from the explicit dimension-correlation mechanism rather than from increased model capacity or the spatial inductive bias of the CNN, the work would supply a concrete architectural alternative to MLP-based interaction functions in neural CF and could encourage further study of dimension-wise interaction maps.

major comments (2)
  1. [Experiments] The central experimental claim (outperformance via explicit pairwise and high-order dimension correlation modeling) rests on comparisons to baselines, yet the manuscript provides no ablation that holds parameter count fixed while replacing the outer-product + CNN structure with an equivalent-capacity MLP or a non-convolutional aggregator on the same 2-D map. Without this control, the reported gains cannot be attributed to the claimed correlation mechanism rather than expressivity.
  2. [Introduction / §3] The abstract and motivation assert that existing neural CF models “do not explicitly consider the correlations among embedding dimensions,” but the manuscript does not supply a formal argument or controlled comparison demonstrating that standard MLP or factorization-machine interaction functions are incapable of learning such correlations when given sufficient capacity.
minor comments (2)
  1. [Experiments] Table captions and axis labels should explicitly state whether reported metrics are averaged over multiple random seeds and whether hyper-parameter search was performed with the same budget for all methods.
  2. [§3] The three instantiations of ConvNCF are described at a high level; a single diagram or pseudocode block showing the precise tensor shapes after the outer product and the CNN filter configuration would improve reproducibility.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback. We address the two major comments point-by-point below, indicating where revisions will be made to strengthen the manuscript.

read point-by-point responses

  1. Referee: [Experiments] The central experimental claim (outperformance via explicit pairwise and high-order dimension correlation modeling) rests on comparisons to baselines, yet the manuscript provides no ablation that holds parameter count fixed while replacing the outer-product + CNN structure with an equivalent-capacity MLP or a non-convolutional aggregator on the same 2-D map. Without this control, the reported gains cannot be attributed to the claimed correlation mechanism rather than expressivity.

    Authors: We agree this is a valid concern. The current experiments compare against published baselines but do not include capacity-controlled ablations against an MLP or non-convolutional aggregator on the interaction map. In the revision we will add such experiments (with parameter counts matched via hidden-layer sizing) to better isolate the contribution of the outer-product + CNN design. revision: yes

  2. Referee: [Introduction / §3] The abstract and motivation assert that existing neural CF models “do not explicitly consider the correlations among embedding dimensions,” but the manuscript does not supply a formal argument or controlled comparison demonstrating that standard MLP or factorization-machine interaction functions are incapable of learning such correlations when given sufficient capacity.

    Authors: The manuscript's wording emphasizes the lack of explicit modeling (via outer product) rather than claiming that MLPs or FMs are theoretically incapable of capturing dimension correlations implicitly. Universal approximation results imply that sufficiently large MLPs can represent such functions. Our contribution is the explicit, structured construction. We will revise the abstract and introduction to replace “do not explicitly consider” with clearer language distinguishing explicit versus implicit modeling and will not add a formal impossibility proof. revision: partial

Circularity Check

0 steps flagged

No circularity: new architecture defined independently and validated on external data

full rationale

The paper proposes ConvNCF as a new neural CF architecture using outer product on embeddings followed by CNN layers. This is an explicit design choice, not derived from prior equations that reduce the claimed benefit to a fitted parameter or self-citation. The justification rests on empirical results against baselines on two datasets, which are independent external benchmarks rather than internal fits. No self-definitional loops, no predictions that are statistically forced by construction, and no load-bearing self-citations appear in the provided text. The derivation chain is self-contained as an architectural proposal.

Axiom & Free-Parameter Ledger

2 free parameters · 2 axioms · 0 invented entities

The central claim rests on the standard definition of outer product and convolutional layers plus the modeling assumption that dimension correlations are under-modeled in prior neural CF. No new physical entities or ad-hoc constants are introduced beyond typical neural hyperparameters.

free parameters (2)
  • embedding dimension
    Standard hyperparameter controlling size of user and item vectors; value not stated in abstract.
  • CNN filter sizes and depths
    Architectural choices that determine how high-order correlations are extracted; not specified in abstract.
axioms (2)
  • standard math Outer product of two vectors produces a matrix whose entries capture all pairwise products of dimensions.
    Invoked when the abstract states that outer product explicitly models pairwise correlations.
  • domain assumption Convolutional layers can extract higher-order patterns from the outer-product matrix.
    Stated as the second design feature in the abstract.

pith-pipeline@v0.9.0 · 5753 in / 1424 out tokens · 24316 ms · 2026-05-25T15:02:56.381303+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

67 extracted references · 67 canonical work pages · 1 internal anchor

  1. [1]

    Ting Bai, Ji-Rong Wen, Jun Zhang, and Wayne Xin Zhao. 2017. A Neural Collaborative Filtering Model with Interaction- based Neighborhood. In CIKM. 1979–1982

    work page 2017

  2. [2]

    Immanuel Bayer, Xiangnan He, Bhargav Kanagal, and Steffen Rendle. 2017. A Generic Coordinate Descent Framework for Learning from Implicit Feedback. In WWW. 1341–1350

    work page 2017

  3. [3]

    James Bennett, Stan Lanning, et al. 2007. The netflix prize. In Proceedings of KDD cup and workshop , Vol. 2007. New York, NY, USA, 35

    work page 2007

  4. [4]

    Alex Beutel, Paul Covington, Sagar Jain, Can Xu, Jia Li, Vince Gatto, and Ed H. Chi. 2018. Latent Cross: Making Use of Context in Recurrent Recommender Systems. In WSDM. 46–54

    work page 2018

  5. [5]

    Da Cao, Xiangnan He, Liqiang Nie, Xiaochi Wei, Xia Hu, Shunxiang Wu, and Tat-Seng Chua. 2017. Cross-Platform App Recommendation by Jointly Modeling Ratings and Texts. ACM Transactions on Information Systems 35, 4 (July 2017), 37:1–37:27

    work page 2017

  6. [6]

    Da Cao, Liqiang Nie, Xiangnan He, Xiaochi Wei, Shunzhi Zhu, and Tat-Seng Chua. 2017. Embedding factorization models for jointly recommending items and user generated lists. In SIGIR. ACM, 585–594

    work page 2017

  7. [7]

    Jingyuan Chen, Hanwang Zhang, Xiangnan He, Liqiang Nie, Wei Liu, and Tat-Seng Chua. 2017. Attentive Collaborative Filtering: Multimedia Recommendation with Item- and Component-Level Attention. In SIGIR. 335–344

    work page 2017

  8. [8]

    Tianqi Chen, Weinan Zhang, Qiuxia Lu, Kailong Chen, Zhao Zheng, and Yong Yu. 2012. SVDFeature: a toolkit for feature-based collaborative filtering. Journal of Machine Learning Research 13, Dec (2012), 3619–3622

    work page 2012

  9. [9]

    Xu Chen, Yongfeng Zhang, Hongteng Xu, Zheng Qin, and Hongyuan Zha. 2018. Adversarial Distillation for Efficient Recommendation with External Knowledge. ACM Transactions on Information Systems (TOIS) 37, 1 (2018), 12

    work page 2018

  10. [10]

    Zhiyong Cheng, Xiaojun Chang, Lei Zhu, Rose C Kanjirathinkal, and Mohan Kankanhalli. 2019. MMALFM: Explainable recommendation by leveraging reviews and images. ACM Transactions on Information Systems (TOIS) 37, 2 (2019), 16

    work page 2019

  11. [11]

    Zhiyong Cheng, Ying Ding, Xiangnan He, Lei Zhu, Xuemeng Song, and Mohan S Kankanhalli. 2018. Aˆ 3NCF: An Adaptive Aspect Attention Model for Rating Prediction.. In IJCAI. 3748–3754

    work page 2018

  12. [12]

    Zhiyong Cheng, Jialie Shen, Lei Zhu, Mohan S Kankanhalli, and Liqiang Nie. 2017. Exploiting Music Play Sequence for Music Recommendation.. In IJCAI. 3654–3660

    work page 2017

  13. [13]

    Paul Covington, Jay Adams, and Emre Sargin. 2016. Deep Neural Networks for YouTube Recommendations. In RecSys. 191–198

    work page 2016

  14. [14]

    George Cybenko. 1989. Approximation by superpositions of a sigmoidal function. Mathematics of control, signals and systems 2, 4 (1989), 303–314

    work page 1989

  15. [15]

    Shuiguang Deng, Longtao Huang, Guandong Xu, Xindong Wu, and Zhaohui Wu. 2017. On deep learning for trust- aware recommendations in social networks. IEEE transactions on neural networks and learning systems 28, 5 (2017), 1164–1177

    work page 2017

  16. [16]

    Jingtao Ding, Fuli Feng, Xiangnan He, Guanghui Yu, Yong Li, and Depeng Jin. 2018. An Improved Sampler for Bayesian Personalized Ranking by Leveraging View Data. In WWW. 13–14

    work page 2018

  17. [17]

    Xiangnan He Xiang Wang Cheng Luo Yiqun Liu Feng, Fuli and Tat-Seng Chua. 2019. Temporal Relational Ranking for Stock Prediction. ACM Transactions on Information Systems (TOIS) 37, 2 (2019), 27

    work page 2019

  18. [18]

    Matt W Gardner and SR Dorling. 1998. Artificial neural networks (the multilayer perceptron)-a review of applications in the atmospheric sciences. Atmospheric environment 32, 14-15 (1998), 2627–2636

    work page 1998

  19. [19]

    Carlos A Gomez-Uribe and Neil Hunt. 2016. The netflix recommender system: Algorithms, business value, and innovation. ACM Transactions on Management Information Systems (TMIS) 6, 4 (2016), 13

    work page 2016

  20. [20]

    Xinyu Guan, Zhiyong Cheng, Xiangnan He, Yongfeng Zhang, Zhibo Zhu, Qinke Peng, and Tat-Seng Chua. 2019. Attentive Aspect Modeling for Review-aware Recommendation. ACM Transactions on Information Systems (TOIS) 37, 3 (2019), 28

    work page 2019

  21. [21]

    Yangyang Guo, Zhiyong Cheng, Liqiang Nie, Yinglong Wang, Jun Ma, and Mohan Kankanhalli. 2019. Attentive Long Short-Term Preference Modeling for Personalized Product Search. ACM Transactions on Information Systems (TOIS) 37, 2 (2019), 19

    work page 2019

  22. [22]

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. 2016. Deep residual learning for image recognition. InCVPR. 770–778

    work page 2016

  23. [23]

    Ruining He and Julian McAuley. 2016. VBPR: Visual Bayesian Personalized Ranking from Implicit Feedback.. In AAAI. 144–150

    work page 2016

  24. [24]

    Xiangnan He and Tat-Seng Chua. 2017. Neural factorization machines for sparse predictive analytics. InSIGIR. 355–364

    work page 2017

  25. [25]

    Xiangnan He, Xiaoyu Du, Xiang Wang, Feng Tian, Jinhui Tang, and Tat-Seng Chua. 2018. Outer Product-based Neural Collaborative Filtering. In IJCAI

    work page 2018

  26. [26]

    Xiangnan He, Zhankui He, Xiaoyu Du, and Tat-Seng Chua. 2018. Adversarial Personalized Ranking for Item Recom- mendation. In SIGIR. ACM Transactions on Information Systems, Vol. 0, No. 0, Article 111. Publication date: 2019. 111:22 Xiaoyu Du, et al

    work page 2018

  27. [27]

    Xiangnan He, Zhenkui He, Jingkuan Song, Zhenguang Liu, Yu-Gang Jiang, and Tat-Seng Chua. 2018. NAIS: Neural Attentive Item Similarity Model for Recommendation. IEEE Transactions on Knowledge and Data Engineering (2018)

    work page 2018

  28. [28]

    Xiangnan He, Lizi Liao, Hanwang Zhang, Liqiang Nie, Xia Hu, and Tat-Seng Chua. 2017. Neural collaborative filtering. In WWW. 173–182

    work page 2017

  29. [29]

    Xiangnan He, Jinhui Tang, Xiaoyu Du, Richang Hong, Tongwei Ren, and Tat-Seng Chua. 2019. Fast Matrix Factorization with Non-Uniform Weights on Missing Data. IEEE transactions on neural networks and learning systems (2019)

    work page 2019

  30. [30]

    Xiangnan He, Hanwang Zhang, Min-Yen Kan, and Tat-Seng Chua. 2016. Fast matrix factorization for online recom- mendation with implicit feedback. In SIGIR. 549–558

    work page 2016

  31. [31]

    Sepp Hochreiter and Jürgen Schmidhuber. 1997. Long short-term memory. Neural computation 9, 8 (1997), 1735–1780

    work page 1997

  32. [32]

    Kurt Hornik. 1991. Approximation capabilities of multilayer feedforward networks. Neural networks 4, 2 (1991), 251–257

    work page 1991

  33. [33]

    Cheng-Kang Hsieh, Longqi Yang, Yin Cui, Tsung-Yi Lin, Serge Belongie, and Deborah Estrin. 2017. Collaborative metric learning. In WWW. 193–201

    work page 2017

  34. [34]

    Gao Huang, Zhuang Liu, Kilian Q Weinberger, and Laurens van der Maaten. 2017. Densely connected convolutional networks. In CVPR. 4700–4708

    work page 2017

  35. [35]

    Santosh Kabbur, Xia Ning, and George Karypis. 2013. Fism: factored item similarity models for top-n recommender systems. In SIGKDD. ACM, 659–667

    work page 2013

  36. [36]

    Donghyun Kim, Chanyoung Park, Jinoh Oh, Sungyoung Lee, and Hwanjo Yu. 2016. Convolutional matrix factorization for document context-aware recommendation. In RecSys. ACM, 233–240

    work page 2016

  37. [37]

    Yehuda Koren. 2008. Factorization meets the neighborhood: a multifaceted collaborative filtering model. In SIGKDD. ACM, 426–434

    work page 2008

  38. [38]

    Yehuda Koren, Robert Bell, and Chris Volinsky. 2009. Matrix factorization techniques for recommender systems. Computer 42, 8 (2009)

    work page 2009

  39. [39]

    Alex Krizhevsky, Ilya Sutskever, and Geoffrey E Hinton. 2012. Imagenet classification with deep convolutional neural networks. In NIPS. 1097–1105

    work page 2012

  40. [40]

    Dawen Liang, Laurent Charlin, James McInerney, and David M Blei. 2016. Modeling user exposure in recommendation. In WWW. 951–961

    work page 2016

  41. [41]

    Zhouhan Lin, Minwei Feng, Cicero Nogueira dos Santos, Mo Yu, Bing Xiang, Bowen Zhou, and Yoshua Bengio. 2017. A structured self-attentive sentence embedding. In ICLR

    work page 2017

  42. [42]

    Zhongqi Lu, Zhicheng Dou, Jianxun Lian, Xing Xie, and Qiang Yang. 2015. Content-Based Collaborative Filtering for News Topic Recommendation.. In AAAI. 217–223

    work page 2015

  43. [43]

    Xin Luo, MengChu Zhou, Shuai Li, Zhuhong You, Yunni Xia, and Qingsheng Zhu. 2016. A nonnegative latent factor model for large-scale sparse matrices in recommender systems via alternating direction method. IEEE transactions on neural networks and learning systems 27, 3 (2016), 579–592

    work page 2016

  44. [44]

    Zhanyu Ma, Yuping Lai, W Bastiaan Kleijn, Yi-Zhe Song, Liang Wang, and Jun Guo. 2018. Variational Bayesian learning for Dirichlet process mixture of inverted Dirichlet distributions in non-Gaussian image feature modeling. IEEE transactions on neural networks and learning systems 99 (2018), 1–15

    work page 2018

  45. [45]

    Zhanyu Ma, Jing-Hao Xue, Arne Leijon, Zheng-Hua Tan, Zhen Yang, and Jun Guo. 2018. Decorrelation of neutral vector variables: Theory and applications. IEEE transactions on neural networks and learning systems 29, 1 (2018), 129–143

    work page 2018

  46. [46]

    Weike Pan, Qiang Yang, Wanling Cai, Yaofeng Chen, Qing Zhang, Xiaogang Peng, and Zhong Ming. 2019. Transfer to rank for heterogeneous one-class collaborative filtering. ACM Transactions on Information Systems (TOIS) 37, 1 (2019), 10

    work page 2019

  47. [47]

    Tieyun Qian, Bei Liu, Quoc Viet Hung Nguyen, and Hongzhi Yin. 2019. Spatiotemporal representation learning for translation-based poi recommendation. ACM Transactions on Information Systems (TOIS) 37, 2 (2019), 18

    work page 2019

  48. [48]

    Bohui Fang Weinan Zhang Ruiming Tang Minzhe Niu Huifeng Guo Yong Yu Qu, Yanru and Xiuqiang He. 2018. Product-Based Neural Networks for User Response Prediction over Multi-Field Categorical Data. ACM Transactions on Information Systems (TOIS) 37, 1 (2018), 5

    work page 2018

  49. [49]

    Steffen Rendle. 2010. Factorization machines. In ICDM. 995–1000

    work page 2010

  50. [50]

    Steffen Rendle, Christoph Freudenthaler, Zeno Gantner, and Lars Schmidt-Thieme. 2009. BPR: Bayesian personalized ranking from implicit feedback. In UAI. 452–461

    work page 2009

  51. [51]

    Amit Sharma, Jake M Hofman, and Duncan J Watts. 2015. Estimating the causal impact of recommendation systems from observational data. In Proceedings of the Sixteenth ACM Conference on Economics and Computation. ACM, 453–470

    work page 2015

  52. [52]

    Yi Tay, Luu Anh Tuan, and Siu Cheung Hui. 2018. Latent Relational Metric Learning via Memory-based Attention for Collaborative Ranking. In WWW. 729–739

    work page 2018

  53. [53]

    Hongwei Wang, Fuzheng Zhang, Xing Xie, and Minyi Guo. 2018. DKN: Deep Knowledge-Aware Network for News Recommendation. In WWW. 1835–1844. ACM Transactions on Information Systems, Vol. 0, No. 0, Article 111. Publication date: 2019. Convolutional NCF 111:23

    work page 2018

  54. [54]

    Suhang Wang, Jiliang Tang, Yilin Wang, and Huan Liu. 2015. Exploring Implicit Hierarchical Structures for Recom- mender Systems. In IJCAI. 1813–1819

    work page 2015

  55. [55]

    Xiang Wang, Xiangnan He, Liqiang Nie, and Tat-Seng Chua. 2017. Item silk road: Recommending items from information domains to social users. In SIGIR. 185–194

    work page 2017

  56. [56]

    Libing Wu, Cong Quan, Chenliang Li, Qian Wang, Bolong Zheng, and Xiangyang Luo. 2019. A context-aware user-item representation learning for item recommendation. ACM Transactions on Information Systems (TOIS) 37, 2 (2019), 22

    work page 2019

  57. [57]

    Yao Wu, Christopher DuBois, Alice X Zheng, and Martin Ester. 2016. Collaborative denoising auto-encoders for top-n recommender systems. In WSDM. ACM, 153–162

    work page 2016

  58. [58]

    Hong-Jian Xue, Xinyu Dai, Jianbing Zhang, Shujian Huang, and Jiajun Chen. 2017. Deep Matrix Factorization Models for Recommender Systems. In IJCAI. 3203–3209

    work page 2017

  59. [59]

    Fisher Yu and Vladlen Koltun. 2015. Multi-scale context aggregation by dilated convolutions. arXiv preprint arXiv:1511.07122 (2015)

    work page internal anchor Pith review Pith/arXiv arXiv 2015

  60. [60]

    Wenhui Yu, Huidi Zhang, Xiangnan He, Xu Chen, Li Xiong, and Zheng Qin. 2018. Aesthetic-based Clothing Recom- mendation. In WWW. 649–658

    work page 2018

  61. [61]

    Fajie Yuan, Guibing Guo, Joemon M Jose, Long Chen, Haitao Yu, and Weinan Zhang. 2016. Lambdafm: learning optimal ranking with factorization machines using lambda surrogates. In CIKM. ACM, 227–236

    work page 2016

  62. [62]

    Fajie Yuan, Alexandros Karatzoglou, Ioannis Arapakis, Joemon M Jose, and Xiangnan He. 2019. A Simple Convolutional Generative Network for Next Item Recommendation. In WSDM

    work page 2019

  63. [63]

    Fajie Yuan, Xin Xin, Xiangnan He, Guibing Guo, Weinan Zhang, Chua Tat-Seng, and Joemon M Jose. 2018. fbgd: Learning embeddings from positive unlabeled data with bgd. (2018)

    work page 2018

  64. [64]

    Yongfeng Zhang, Qingyao Ai, Xu Chen, and W Bruce Croft. 2017. Joint representation learning for top-n recommen- dation with heterogeneous information sources. In CIKM. 1449–1458

    work page 2017

  65. [65]

    Yongfeng Zhang, Min Zhang, Yiqun Liu, Shaoping Ma, and Shi Feng. 2013. Localized matrix factorization for recommendation based on matrix block diagonal forms. In Proceedings of the 22nd international conference on World Wide Web. ACM, 1511–1520

    work page 2013

  66. [66]

    Wayne Xin Zhao, Wenhui Zhang, Yulan He, Xing Xie, and Ji-Rong Wen. 2018. Automatically learning topics and difficulty levels of problems in online judge systems. ACM Transactions on Information Systems (TOIS) 36, 3 (2018), 27

    work page 2018

  67. [67]

    Zhou Zhao, Hanqing Lu, Deng Cai, Xiaofei He, and Yueting Zhuang. 2016. User preference learning for online social recommendation. IEEE Transactions on Knowledge and Data Engineering 28, 9 (2016), 2522–2534. ACM Transactions on Information Systems, Vol. 0, No. 0, Article 111. Publication date: 2019

    work page 2016