Pith. sign in

REVIEW 3 major objections 5 minor 62 references

Shapley Value-driven Data Pruning for Recommender Systems

T0 review · 3 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read This paper claims that an interaction's training utility can be measured by its Shapley value under a loss-reduction value function, and that pruning the bottom fifth of interactions by this value improves recommendation accuracy and…

desk verdict The pruning gains may be real, but the paper's central equivalence between Eq. 16 and Eq. 17 is false, so the Shapley values do not measure training-loss contribution, and the experimental tables are internally inconsistent. read the letter →

arxiv 2505.22057 v1 pith:FXGDEFHV submitted 2025-05-28 cs.IR

Yansen Zhang , Xiaokun Zhang , Ziqiang Cui , Chen Ma This is my paper
classification cs.IR
keywords ShapleyvaluesdatavaluationpruningrecommendersystemsdenoisingimplicitfeedbackautoencoderFastSHAP
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

Recommender systems are trained on interactions that include accidental clicks and popularity bias, yet some interactions that look noisy still help the model learn, while some clean-looking interactions add little. This paper proposes Shapley Value-driven Valuation (SVV), which scores each observed user–item interaction by its Shapley value—the average marginal contribution of that interaction to reducing the training loss of a denoising autoencoder—and then prunes the lowest-scoring interactions before retraining. On four real-world datasets, dropping the bottom fifth of interactions by this value improves accuracy over the base model by 0.75 to 7.22 percent and outperforms existing intent-based denoising methods. The paper also introduces a simulated-noise injection protocol that provides a verifiable ground truth for measuring how well a denoiser identifies harmful interactions.

What carries the argument

The machinery has three parts. First, a denoising autoencoder (DAE) serves as the base model: it accepts a user's binary interaction vector as input, masks entries during training, and reconstructs the full vector, so it can score subsets of interactions without retraining. Second, FastSHAP is a real-time amortized estimator that trains a neural network to output Shapley values in a single forward pass, using a Shapley-weighted least-squares loss and an additive efficiency normalization so that the outputs converge to exact Shapley values. Third, the value function maps a subset of a user's interactions to the average predicted score over the observed interactions (Eq. 17), which the paper treats as equivalent to negative normalized squared reconstruction error (Eq. 16) for binary feedback. The computed Shapley values then feed a threshold rule that keeps the top 80 percent of interactions and drops the rest.

What would settle it

Take any user vector with observed interactions and distinct predicted scores; the two expressions differ by $(1/m)\sum_v (f_v - f_v^2) - 1$ over observed items, so the claimed equivalence can be checked numerically on a single example. Then rerun the SVV pipeline with the loss-based value function of Eq. 16 in place of the average-score form of Eq. 17 and compare the pruned sets and final metrics; if the rankings change materially, the simplified value function is carrying the result.

Watch

Extended reading notes

Core claim

The paper's central claim is that an interaction's training utility is a measurable, model-driven quantity: its Shapley value under a value function that reflects loss reduction, estimated for every interaction in one forward pass using a real-time estimator. For binary implicit feedback, the value function is presented in simplified form as the average predicted score over a user's observed interactions, taken to be equivalent to the normalized squared reconstruction error. SVV ranks all observed interactions by this value, removes the bottom 20 percent, retrains the base denoising autoencoder on the remainder, and reports consistent gains in Recall@K and NDCG@K across the Ta Feng, CDs, Games, and Movies datasets. The paper interprets these results as showing that pruning should follow training utility rather than assumptions about user intent, and that the computed scores are meaningful enough to identify injected noise and to expose cases where a highly rated item has low training value.

Load-bearing premise

The load-bearing premise is that, for binary implicit feedback, the average predicted score over observed interactions equals the negative normalized squared reconstruction error, so that Shapley values of the simplified value function really are Shapley values of training-loss reduction.

Editorial extensions

If this is right

  • Data denoising in recommender systems can be reframed as utility valuation, so interactions that deviate from assumed intent but help training are retained rather than discarded.
  • The pruning rule needs no external labels or auxiliary signals, only the base model's training signal, so it applies to any masked-input recommender architecture.
  • Because Shapley values are additive and per-interaction, SVV offers an interpretable score that practitioners can inspect, such as a rating-5 item assigned a low training value.
  • The simulated-noise protocol provides a benchmark for comparing denoisers by known noise recovery, while the results show that higher noise recovery does not automatically mean better recommendation performance.
  • The method's gains depend on noise type: it is strongest for unpopular-item noise and can be negative for popular-item noise, so the type of noise determines whether pruning helps.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A testable extension is to recompute the Shapley values with the exact loss-based value function of Eq. 16 instead of the simplified average-score form; if the pruned sets are stable, the simplification is not the active ingredient, and if they shift, the paper's stated equivalence would need revision.
  • The same valuation machinery could be run iteratively, scoring interactions again after each round of pruning, which would turn SVV into an online data-selection loop rather than a one-off filter.
  • The paper's noise-type results suggest that SVV may systematically favor long-tail interactions, so a natural follow-up is to measure catalogue coverage or diversity after pruning and test whether the gains come partly from popularity correction.
  • The benchmark protocol could be reused to evaluate other value functions, for example scoring by marginal contribution to held-out ranking metrics rather than reconstruction loss, to see which definition of training utility best predicts downstream accuracy.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper proposes Shapley Value-driven Valuation (SVV), a data-pruning framework for implicit-feedback recommender systems. SVV trains a denoising autoencoder (DAE), estimates Shapley values of observed user–item interactions with FastSHAP using a value function meant to measure contribution to training-loss reduction, prunes the lowest-valued interactions, and retrains the DAE on the pruned set. The authors also introduce a simulated noise injection protocol and report accuracy and robustness gains over several baselines on four datasets. The central theoretical claim is that, for binary feedback, the negative normalized squared reconstruction error in Eq. (16) is equivalent to the average predicted score in Eq. (17), so that Shapley values computed from the latter quantify each interaction's contribution to reducing training loss.

Significance. If the central derivation were correct, the paper would make a useful contribution: it proposes a principled, model-driven alternative to intent-based denoising, makes code available, and includes a controlled simulated-noise benchmark. The interpretability analysis and the observation that low-rated or 'noisy' interactions can aid training are also valuable. However, the load-bearing equivalence between Eqs. (16) and (17) is algebraically false, so the computed Shapley values do not measure training-loss contribution as claimed. The empirical gains may still be real, but they do not validate the paper's stated mechanism. The paper is therefore not acceptable in its current form.

major comments (3)
  1. [Sec. 4.2, Eqs. (16)–(17)] The claimed equivalence between the negative normalized squared reconstruction error and the average predicted score is false for binary implicit feedback. Expanding Eq. (16) over observed entries with binary r gives v1(S) = (1/m)(2Σ f_v − Σ f_v^2) − 1, where m = 1^T r and the sums run over observed entries, whereas Eq. (17) is v2(S) = (1/m)Σ f_v. The difference is v2(S) − v1(S) = 1 + (1/m)Σ(f_v^2 − f_v). Since f(r_S) depends on the coalition S, this difference is not an additive constant; it changes the marginal contributions of individual interactions. Consequently, the two value functions define different cooperative games and yield different Shapley values. The paper's central interpretation—that low-SVV interactions contribute least to training-loss reduction—does not follow from the method as implemented.
  2. [Sec. 5.5, Table 4] The robustness claim in the abstract and conclusion is not supported by the results for popular-item noise. In Table 4, SVV is worse than the base model on multiple metrics under Popular injected noise, for example CDs R@20 (−3.90%), CDs R@10 (−1.42%), Games R@5 (−2.70%), and Movies R@20 (+0.11%, effectively no gain). The text acknowledges deterioration for popular items but still summarizes the method as improving 'both accuracy and robustness.' This overstates the empirical evidence and should be revised or the robustness claim qualified to the tested noise types.
  3. [Sec. 4.3, Eq. (20)] There is a mismatch between the noise injection rate and the pruning ratio that weakens the interpretation of the noise-detection results. The injection protocol adds k% of each user's original interaction count as noise, so the corrupted set has |D+| + k%|D+| interactions; pruning the bottom k% of the corrupted set removes k% of the corrupted data, not k% of the original data, and the injected noise constitutes k/(100+k)% of the corrupted set. The paper sets both k=20 without explicitly addressing this distinction, so the claim that pruning exactly removes the injected noise is not internally consistent. This affects the interpretation of the 'overlap' percentages in Table 3 and the validation protocol in Sec. 5.3.
minor comments (5)
  1. [Sec. 4.2, sentence after Eq. (17)] The clause 'where minimizing squared error loss closely aligns with maximizing predicted scores' should be replaced by the exact algebraic relationship, since as written it masks the incorrect equivalence in Eqs. (16)–(17).
  2. [Sec. 5.1.4] The DAE corruption (masking) ratio is not specified; the text only says the binary input vector is randomly masked using PyTorch's randint function. This is a key hyperparameter for the base model and should be reported.
  3. [Sec. 5.3, Table 3] The notation 'Top↓' and 'Bottom↑' is not explained. It appears that lower Top overlap and higher Bottom overlap are desirable for detecting injected noise, but the arrow convention should be defined in the caption or text.
  4. [Figure 1 and Figure 2] Figure 1 contains the typo 'Gamses' for 'Games,' and Figure 2 does not label the axes or indicate error bars, which makes the monotonicity claims in Sec. 5.4 difficult to verify.
  5. [Sec. 5.2, Table 2] The paper claims statistical significance via a paired t-test at the 0.05 level but reports only averaged metrics over five runs; no variances, standard errors, or p-values are provided, so the significance claim cannot be checked.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: SVV's Shapley value computation is model-driven and benchmarked externally; the disputed Eq. 16-17 equivalence is a mathematical error, not a circular reduction.

full rationale

The central claim is that Shapley values computed from Eq. 17 quantify marginal contributions to training loss reduction. Section 4.2 states that Eq. 16 defines the value function as the negative normalized squared reconstruction error and claims that for binary implicit feedback r it simplifies to Eq. 17, the average predicted score. These two value functions are not equal: expanding Eq. 16 over observed entries gives v1(S) = (1/m)(2 sum f_v - sum f_v^2) - 1, while Eq. 17 is v2(S) = (1/m) sum f_v; the difference v2 - v1 = 1 + (1/m) sum (f_v^2 - f_v) depends on S, so the cooperative games differ and the Shapley values of Eq. 17 are not the Shapley values of Eq. 16. This invalidates the "training loss reduction" interpretation, but it is a derivation error rather than a circularity: Eq. 17 is not defined in terms of the pruning outcome, no parameter is fitted to the validation benchmark, and no load-bearing step is justified solely by a self-citation. FastSHAP's convergence theorem is cited to external authors [25]. The empirical gains are measured against external baselines on four public datasets, and the noise-overlap analysis uses injected labels only as an evaluation target, not as training signal. The Section 5.4 value-function consistency curves are in-sample by construction, but the accompanying Recall/NDCG curves provide independent grounding. Self-citations [18, 27, 52] occur only in related-work discussion and are not load-bearing. Hence the circularity score is 0, with the caveat that the false Eq. 16-17 equivalence is a serious correctness issue.

Assumptions & free parameters 4 free parameters · 4 assumptions · 0 invented entities

The central claim rests on the value function equivalence, which is false, and on assumptions about FastSHAP accuracy and the simulated-noise protocol. Free hyperparameters such as k and rho are tuned on validation and directly shape the Shapley values. No new entities are introduced.

free parameters (4)
  • pruning/noise ratio k = 20
    Fraction of interactions pruned and fraction of non-interactions injected as noise. Fixed at 20 percent across all datasets; the paper states changing k did not affect conclusions but does not report a sensitivity table.
  • confidence weight rho = selected from {5,10,15,20,25}
    Hyperparameter weighting positive interactions in the DAE loss (Eq. 7); tuned on validation per dataset and directly shapes the gradients and Shapley values.
  • FastSHAP learning rate = selected from {0.001,0.01,0.1}
    Tuned on validation; affects the quality of the amortized Shapley estimator.
  • DAE architecture and hyperparameters = H1=50, lambda_Theta=0.001, LR=0.001, epochs=200
    Standard settings chosen by the authors; they affect the trained model and therefore the Shapley values.
assumptions (4)
  • ad hoc to paper The value function v_{r,y}(S) = (1^T (f(r_S) * r)) / (1^T r) is equivalent to the negative normalized squared reconstruction loss - (1/(1^T r)) || f(r_S) * r - r ||^2_2 for binary r.
    Stated in Sec. 4.2 as a simplification, but the two expressions differ by the term (1/m) times the sum of (a_v - a_v^2) minus 1 for observed entries, so they are not algebraically equivalent unless predictions are binary.
  • domain assumption FastSHAP's amortized estimator converges to exact Shapley values for the chosen value function.
    The paper relies on the universal approximation and convergence results of Jethani et al. (2022) without verifying estimation error on recommender data.
  • domain assumption The DAE, trained with masked inputs and a confidence-weighted squared loss, is an appropriate base model whose reconstruction objective reflects training utility.
    The paper selects DAE because it can handle missing features, but assumes its training loss is the right measure of interaction value for the downstream recommendation task.
  • domain assumption Simulated noise injection, randomly flipping non-interactions to interactions, creates a valid proxy for real-world noise.
    Used as a self-validating benchmark; the paper acknowledges injected noise may not perfectly represent true noise labels.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Shapley Value-driven Data Pruning for Recommender Systems." pith.science (2026). https://pith.science/paper/FXGDEFHV

@misc{pith2026250522057,
  author       = {Pith},
  title        = {Pith review of: Shapley Value-driven Data Pruning for Recommender Systems},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FXGDEFHV}},
  note         = {Machine review of arXiv:2505.22057}
}
read the original abstract

Recommender systems often suffer from noisy interactions like accidental clicks or popularity bias. Existing denoising methods typically identify users' intent in their interactions, and filter out noisy interactions that deviate from the assumed intent. However, they ignore that interactions deemed noisy could still aid model training, while some ``clean'' interactions offer little learning value. To bridge this gap, we propose Shapley Value-driven Valuation (SVV), a framework that evaluates interactions based on their objective impact on model training rather than subjective intent assumptions. In SVV, a real-time Shapley value estimation method is devised to quantify each interaction's value based on its contribution to reducing training loss. Afterward, SVV highlights the interactions with high values while downplaying low ones to achieve effective data pruning for recommender systems. In addition, we develop a simulated noise protocol to examine the performance of various denoising approaches systematically. Experiments on four real-world datasets show that SVV outperforms existing denoising methods in both accuracy and robustness. Further analysis also demonstrates that our SVV can preserve training-critical interactions and offer interpretable noise assessment. This work shifts denoising from heuristic filtering to principled, model-driven interaction valuation.

Figures

Figures reproduced from arXiv: 2505.22057 by the authors.

Figure 1
Figure 1. Performance comparisons of the Clean (all rating [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Contrasting Shapley value-based exclusion strate [PITH_FULL_IMAGE:figures/full_fig_p009_2.png] view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

62 extracted references · 49 canonical work pages

  1. [27]

    Renqi Jia, Xiaokun Zhang, Bowei He, Qiannan Zhu, Weitao Xu, Jiehao Chen, and Chen Ma. 2025. Beyond Models! Explainable Data Valuation and Metric Adaption for Recommendation. InSDM. 203–212

    work page 2025

  2. [18]

    Bowei He and Chen Ma. 2024. Interpretable Triplet Importance for Personalized Ranking. InCIKM. 809–818

    2024

  3. [1]

    Liat Antwarg, Ronnie Mindlin Miller, Bracha Shapira, and Lior Rokach. 2021. Explaining anomalies detected by autoencoders using Shapley Additive Explana- tions.Expert systems with applications186 (2021), 115736. KDD ’25, August 3–7, 2025, Toronto, ON, Canada. Yansen Zhang, Xiaokun Zhang, Ziqiang Cui, and Chen Ma

    work page 2021

  4. [2]

    Zhi Bian, Shaojun Zhou, Hao Fu, Qihong Yang, Zhenqi Sun, Junjie Tang, Guiquan Liu, Kaikui Liu, and Xiaolong Li. 2021. Denoising user-aware memory network for recommendation. InRecSys. 400–410

    work page 2021

  5. [3]

    Raghavendra Chalapathy and Sanjay Chawla. 2019. Deep learning for anomaly detection: A survey.arXiv preprint arXiv:1901.03407(2019)

    arXiv 2019

  6. [4]

    Hugh Chen, Ian C Covert, Scott M Lundberg, and Su-In Lee. 2023. Algorithms to estimate Shapley value feature attributions.Nat. Mac. Intell.5 (2023), 590–601

    work page 2023

  7. [5]

    Huiyuan Chen, Yusan Lin, Menghai Pan, Lan Wang, Chin-Chia Michael Yeh, Xiaoting Li, Yan Zheng, Fei Wang, and Hao Yang. 2022. Denoising self-attentive sequential recommendation. InRecSys. 92–101

    work page 2022

  8. [6]

    Jiawei Chen, Hande Dong, Xiang Wang, Fuli Feng, Meng Wang, and Xiangnan He. 2023. Bias and debias in recommender system: A survey and future directions. TOIS41, 3 (2023), 1–39

    work page 2023

Show all 62 references
  1. [7]

    Lu Chen, Siyu Lou, Keyan Zhang, Jin Huang, and Quanshi Zhang. 2023. Harsanyinet: Computing accurate shapley values in a single forward propagation. arXiv preprint arXiv:2304.01811(2023)

    2023 arXiv

  2. [8]

    Dennis Cook

    R. Dennis Cook. 2000. Detection of Influential Observation in Linear Regression. Technometrics42, 1 (2000), 65–68

    2000

  3. [9]

    Ian Covert and Su-In Lee. 2021. Improving kernelshap: Practical shapley value estimation using linear regression. InAISTATS. 3457–3465

    2021

  4. [10]

    Jingtao Ding, Guanghui Yu, Xiangnan He, Fuli Feng, Yong Li, and Depeng Jin

  5. [11]

    Steve Fox, Kuldeep Karnawat, Mark Mydland, Susan Dumais, and Thomas White

  6. [12]

    Zeno Gantner, Lucas Drumond, Christoph Freudenthaler, and Lars Schmidt- Thieme. 2012. Personalized ranking for non-uniformly sampled items. InKDD. 231–247

    2012

  7. [13]

    Yunjun Gao, Yuntao Du, Yujia Hu, Lu Chen, Xinjun Zhu, Ziquan Fang, and Baihua Zheng. 2022. Self-guided learning to denoise for robust recommendation. In SIGIR. 1412–1422

    2022

  8. [14]

    Yingqiang Ge, Mostafa Rahmani, Athirai Irissappane, Jose Sepulveda, James Caverlee, and Fei Wang. 2023. Automated data denoising for recommendation. arXiv preprint arXiv:2305.07070(2023)

    2023 arXiv

  9. [15]

    Amirata Ghorbani and James Y. Zou. 2019. Data Shapley: Equitable Valuation of Data for Machine Learning. InICML, Vol. 97. 2242–2251

    2019

  10. [16]

    Xavier Glorot and Yoshua Bengio. 2010. Understanding the difficulty of training deep feedforward neural networks. InAISTATS. 249–256

    2010

  11. [17]

    Stefan Grafberger, Paul Groth, Julia Stoyanovich, and Sebastian Schelter. 2022. Data distribution debugging in machine learning pipelines.VLDBJ31, 5 (2022), 1103–1126

    2022

  12. [19]

    Xiangnan He, Kuan Deng, Xiang Wang, Yan Li, Yongdong Zhang, and Meng Wang. 2020. Lightgcn: Simplifying and powering graph convolution network for recommendation. InSIGIR. 639–648

    2020

  13. [20]

    Zhuangzhuang He, Yifan Wang, Yonghui Yang, Peijie Sun, Le Wu, Haoyue Bai, Jinqi Gong, Richang Hong, and Min Zhang. 2024. Double correction framework for denoising recommendation. InKDD. 1062–1072

    2024

  14. [21]

    Chun-Nan Hsu, Hao-Hsiang Chung, and Han-Shen Huang. 2004. Mining skewed and sparse transaction data for personalized shopping recommendation.Machine Learning57 (2004), 35–59

    2004

  15. [22]

    Yifan Hu, Yehuda Koren, and Chris Volinsky. 2008. Collaborative filtering for implicit feedback datasets. InICDM. 263–272

    2008

  16. [23]

    Junjie Huang, Ruobing Xie, Qi Cao, Huawei Shen, Shaoliang Zhang, Feng Xia, and Xueqi Cheng. 2023. Negative can be positive: Signed graph neural networks for recommendation.IPM60, 4 (2023), 103403

    2023

  17. [24]

    Jakub Jakubowski, Przemysław Stanisz, Szymon Bobek, and Grzegorz J Nalepa

  18. [25]

    Neil Jethani, Mukund Sudarshan, Ian Covert, Su-In Lee, and Rajesh Ranganath

  19. [26]

    Spanos, and Dawn Song

    Ruoxi Jia, David Dao, Boxin Wang, Frances Ann Hubis, Nezihe Merve Gürel, Bo Li, Ce Zhang, Costas J. Spanos, and Dawn Song. 2019. Efficient Task-Specific Data Valuation for Nearest Neighbor Algorithms.Proc. VLDB Endow.12, 11 (2019), 1610–1623

    2019

  20. [28]

    Bojan Karlaš, David Dao, Matteo Interlandi, Bo Li, Sebastian Schelter, Wentao Wu, and Ce Zhang. 2022. Data debugging with shapley importance over end-to-end machine learning pipelines.arXiv preprint arXiv:2204.11131(2022)

    2022 arXiv

  21. [29]

    Youngho Kim, Ahmed Hassan, Ryen W White, and Imed Zitouni. 2014. Modeling dwell time to predict click-level satisfaction. InWSDM. 193–202

    2014

  22. [30]

    Diederik P Kingma and Jimmy Ba. 2014. Adam: A method for stochastic opti- mization.arXiv preprint arXiv:1412.6980(2014)

    2014 arXiv

  23. [31]

    Dongha Lee, SeongKu Kang, Hyunjun Ju, Chanyoung Park, and Hwanjo Yu. 2021. Bootstrapping user and item representations for one-class collaborative filtering. InSIGIR. 317–326

    2021

  24. [32]

    Zhong Li, Yuxuan Zhu, and Matthijs Van Leeuwen. 2023. A survey on explainable anomaly detection.TKDD18, 1 (2023), 1–54

    2023

  25. [33]

    Weilin Lin, Xiangyu Zhao, Yejing Wang, Yuanshao Zhu, and Wanyu Wang. 2023. Autodenoise: Automatic data instance denoising for recommendations. InWWW. 1003–1011

    2023

  26. [34]

    Hongyu Lu, Min Zhang, and Shaoping Ma. 2018. Between clicks and satisfaction: Study on multi-phase user preferences and satisfaction for online news reading. InSIGIR. 435–444

    2018

  27. [35]

    Scott M Lundberg and Su-In Lee. 2017. A unified approach to interpreting model predictions. InNIPS. 4768–4777

    2017

  28. [36]

    Xiaoxiao Ma, Jia Wu, Shan Xue, Jian Yang, Chuan Zhou, Quan Z Sheng, Hui Xiong, and Leman Akoglu. 2021. A comprehensive survey on graph anomaly detection with deep learning.TKDE35, 12 (2021), 12012–12038

    2021

  29. [37]

    Jianmo Ni, Jiacheng Li, and Julian McAuley. 2019. Justifying recommendations using distantly-labeled reviews and fine-grained aspects. InEMNLP-IJCNLP. 188–197

    2019

  30. [38]

    Steffen Rendle, Christoph Freudenthaler, Zeno Gantner, and Lars Schmidt-Thieme

  31. [39]

    Luis M Ruiz, Federico Valenciano, and Jose M Zarzuelo. 1998. The family of least square values for transferable utility games.Games and Economic Behavior24, 1-2 (1998), 109–130

    1998

  32. [40]

    Suvash Sedhain, Aditya Krishna Menon, Scott Sanner, and Lexing Xie. 2015. Autorec: Autoencoders meet collaborative filtering. InWWW. 111–112

    2015

  33. [41]

    Lloyd S Shapley et al. 1953. A value for n-person games. (1953)

    1953

  34. [42]

    Erik Štrumbelj and Igor Kononenko. 2014. Explaining prediction models and individual predictions with feature contributions.Knowl. Inf. Syst.41 (2014), 647–665

    2014

  35. [43]

    Mukund Sundararajan and Amir Najmi. 2020. The many Shapley values for model explanation. InICML. 9269–9278

    2020

  36. [44]

    Tianle Wang, Lianghao Xia, and Chao Huang. 2023. Denoised self-augmented learning for social recommendation.arXiv preprint arXiv:2305.12685(2023)

    2023 arXiv

  37. [45]

    Wenjie Wang, Fuli Feng, Xiangnan He, Liqiang Nie, and Tat-Seng Chua. 2021. Denoising implicit feedback for recommendation. InWSDM. 373–381

    2021

  38. [46]

    Yu Wang, Xin Xin, Zaiqiao Meng, Joemon M Jose, Fuli Feng, and Xiangnan He

  39. [47]

    Zongwei Wang, Min Gao, Wentao Li, Junliang Yu, Linxin Guo, and Hongzhi Yin. 2023. Efficient bi-level optimization for recommendation denoising. InKDD. 2502–2511

    2023

  40. [48]

    Yao Wu, Christopher DuBois, Alice X Zheng, and Martin Ester. 2016. Collab- orative denoising auto-encoders for top-n recommender systems. InWSDM. 153–162

    2016

  41. [49]

    Feng Xue, Xiangnan He, Xiang Wang, Jiandong Xu, Kai Liu, and Richang Hong

  42. [50]

    Tom Yan and Ariel D Procaccia. 2021. If you like shapley then you’ll love the core. InAAAI, Vol. 35. 5751–5759

    2021

  43. [51]

    Haibo Ye, Xinjie Li, Yuan Yao, and Hanghang Tong. 2023. Towards robust neural graph collaborative filtering via structure denoising and embedding perturbation. TOIS41, 3 (2023), 1–28

    2023

  44. [52]

    Learning robust recommenders through cross-model agreement. InWWW. 2015–2025

    2015

  45. [53]

    Yansen Zhang, Chenhao Hu, Genan Dai, Weiyang Kong, and Yubao Liu. 2021. Self- adaptive Graph Neural Networks for Personalized Sequential Recommendation. InICONIP. 608–619

    2021

  46. [54]

    Haiyuan Zhao, Lei Zhang, Jun Xu, Guohao Cai, Zhenhua Dong, and Ji-Rong Wen

  47. [56]

    Deep item-based collaborative filtering for top-n recommendation.TOIS 37, 3 (2019), 1–25

    2019

  48. [59]

    Yansen Zhang, Bowei He, Xiaokun Zhang, Haolun Wu, Zexu Sun, and Chen Ma. 2025. Counterfactual Multi-player Bandits for Explainable Recommendation Diversification.arXiv preprint arXiv:2505.21165(2025)

    2025 arXiv

  49. [2005]

    Evaluating implicit measures to improve web search.TOIS23, 2 (2005), 147–168

    2005

  50. [2009]

    BPR: Bayesian Personalized Ranking from Implicit Feedback. InUAI. 452– 461

  51. [2019]

    TKDE33, 2 (2019), 667–681

    Sampler design for bayesian personalized ranking by leveraging view data. TKDE33, 2 (2019), 667–681

    2019

  52. [2021]

    Anomaly Detection in Asset Degradation Process Using Variational Au- toencoder and Explanations.Sensors22, 1 (2021), 291

    2021

  53. [2022]

    FastSHAP: Real-Time Shapley Value Estimation. InICLR

  54. [2023]

    InRecSys

    Uncovering User Interest from Biased and Noised Watch Time in Video Recommendation. InRecSys. 528–539

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.