pith. sign in

arxiv: 2605.20721 · v1 · pith:3G65JD7Dnew · submitted 2026-05-20 · 💻 cs.LG

Robust Recommendation from Noisy Implicit Feedback: A GMM-Weighted Bayes-label Transition Matrix Framework

Zongyu Li , Xuanyu Liu , Gongce Cao , Shirui Sun , Yaqi Fang , Yongshuai Yu This is my paper

Pith reviewed 2026-05-21 05:45 UTC · model grok-4.3

classification 💻 cs.LG
keywords noisy implicit feedbackrecommender systemsBayes-label transition matrixGaussian Mixture Modelrobust recommendationlabel noisedenoising methods
0
0 comments X

The pith

A GMM assigns instance-specific reliability scores that calibrate a Bayes-label transition matrix, letting recommender systems use every noisy implicit feedback sample while cutting estimation variance.

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

The paper introduces RGBT, a framework that keeps all available implicit feedback instead of discarding noisy instances. It fits a Gaussian Mixture Model to the data to produce per-instance reliability scores, then uses those scores to adjust the entries of a Bayes-label transition matrix. The adjusted matrix yields consistent parameter estimates with lower variance than either sample-dropping methods or uncalibrated transition-matrix approaches. If the calibration works as described, systems can train on the full noisy dataset without the bias that previously limited transition-matrix techniques.

Core claim

By deriving instance-specific reliability scores from a Gaussian Mixture Model and using them to calibrate the Bayes-label transition matrix, the RGBT framework simultaneously achieves full sample utilization, consistent estimation, and a significant reduction in estimation variance for learning from noisy implicit feedback.

What carries the argument

GMM-weighted calibration of the Bayes-label transition matrix, in which a Gaussian Mixture Model supplies per-instance reliability weights that adjust the transition probabilities to reduce bias and variance.

If this is right

  • Recommender models can train on the entire set of implicit feedback without any instance being discarded.
  • The calibrated transition matrix produces estimates with both lower bias and lower variance than prior BLTM methods.
  • Empirical performance on real-world and label-flipped datasets exceeds both reliable-sample denoising baselines and uncalibrated transition-matrix baselines.

Where Pith is reading between the lines

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

  • The same GMM-calibration idea could be tested on noisy-label problems outside recommendation, such as image classification with web-scraped tags.
  • If the reliability scores prove stable across different recommendation domains, the framework might reduce the engineering effort spent on manual data cleaning pipelines.
  • A natural next measurement is whether the variance reduction persists when the underlying model is a deep neural network rather than a simpler matrix-factorization form.

Load-bearing premise

A Gaussian Mixture Model fitted to the observed data can generate instance-specific reliability scores that correctly adjust the transition matrix without introducing fresh bias or model mismatch for typical recommendation noise.

What would settle it

Run RGBT on a synthetic dataset whose noise distribution is deliberately non-Gaussian and check whether the claimed reduction in estimation variance and consistency disappear while bias reappears.

Figures

Figures reproduced from arXiv: 2605.20721 by Gongce Cao, Shirui Sun, Xuanyu Liu, Yaqi Fang, Yongshuai Yu, Zongyu Li.

Figure 1
Figure 1. Figure 1: Performance Shift under Noise Increase [PITH_FULL_IMAGE:figures/full_fig_p011_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Parameter sensitivity analysis of λ and ρ on ML-100k dataset. Left column shows impact of λ (log-scale), Right column shows impact of ρ. 13 [PITH_FULL_IMAGE:figures/full_fig_p013_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Performance comparison of iterative strategy vs. fixed thresholds ( [PITH_FULL_IMAGE:figures/full_fig_p014_3.png] view at source ↗
read the original abstract

Learning from implicit feedback in recommender systems is fundamentally challenged by pervasive label noise. While conventional denoising approaches often discard noisy instances to ensure robustness, this strategy inevitably suffers from low data utilization. Alternative methods that employ a Bayes-label transition matrix (BLTM) can leverage all available data, but their estimates tend to be biased in practical recommendation scenarios. To address these limitations, this paper proposes a Robust GMM-weighted Bayes-label Transition Matrix framework (RGBT). Our solution utilizes a Gaussian Mixture Model (GMM) to derive instance-specific reliability scores, which systematically calibrate the BLTM estimation to mitigate bias. Theoretical analysis confirms that our approach, by leveraging the BLTM framework with GMM calibration, simultaneously ensures full sample utilization, delivers consistent estimation, and critically, achieves a significant reduction in estimation variance. Extensive experiments on multiple real-world and synthetically flipped datasets demonstrate that RGBT not only utilizes noisy samples more effectively than mainstream reliable sample-based denoising methods, but also achieves significantly superior calibration capability of the transition matrix compared to state-of-the-art transition matrix-based denoising approaches.

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 paper proposes the RGBT framework for robust recommendation from noisy implicit feedback. It employs a Gaussian Mixture Model (GMM) to compute instance-specific reliability scores that calibrate a Bayes-label transition matrix (BLTM), aiming to achieve full sample utilization, consistent parameter estimation, and reduced estimation variance relative to prior denoising approaches. Theoretical analysis is claimed to establish these properties, with supporting experiments on real-world and synthetically flipped datasets showing improved performance over reliable-sample and transition-matrix baselines.

Significance. If the GMM calibration step can be shown to recover accurate per-instance reliability weights without material misspecification bias under realistic implicit-feedback noise, the result would be significant: it would resolve the data-utilization versus robustness trade-off that currently forces practitioners to discard noisy samples. The explicit variance-reduction guarantee alongside consistency is a strong point, as is the use of all samples rather than filtering.

major comments (2)
  1. [§4] §4 (Theoretical Analysis): The consistency and variance-reduction claims rest on the premise that GMM-derived reliability scores correctly calibrate the BLTM without introducing systematic bias. No error bounds or sensitivity analysis are provided for GMM misspecification under typical implicit-feedback noise sources (position bias, selection effects, heterogeneous user behavior), which directly undermines the central theoretical guarantees.
  2. [§3.2] §3.2 (GMM-weighted Calibration): The procedure for fitting the GMM and extracting instance-specific weights is described at a high level; it is unclear whether the number of components, initialization, or convergence criteria are chosen independently of the downstream recommendation objective. If these choices are tuned on the same noisy data used for BLTM estimation, the claimed consistency may be circular.
minor comments (2)
  1. [Experiments] The experimental section should explicitly state the synthetic noise-generation process (flipping probability, which labels are flipped) and include an ablation on GMM component count to demonstrate robustness.
  2. Notation for the reliability weight w_i and the calibrated transition matrix should be introduced once in the preliminaries and used consistently thereafter to improve readability.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the detailed and constructive feedback on our manuscript. We address each major comment below with clarifications and indicate planned revisions where appropriate.

read point-by-point responses

  1. Referee: [§4] §4 (Theoretical Analysis): The consistency and variance-reduction claims rest on the premise that GMM-derived reliability scores correctly calibrate the BLTM without introducing systematic bias. No error bounds or sensitivity analysis are provided for GMM misspecification under typical implicit-feedback noise sources (position bias, selection effects, heterogeneous user behavior), which directly undermines the central theoretical guarantees.

    Authors: We appreciate the referee highlighting this foundational assumption. Section 4 derives consistency and variance reduction conditionally on the GMM providing reliable per-instance weights that correctly calibrate the BLTM. The current manuscript does not include explicit error bounds or a formal sensitivity analysis for GMM misspecification. However, the synthetic experiments deliberately introduce controlled flips that emulate position bias, selection effects, and heterogeneous noise, and the real-world results remain stable across datasets. In revision we will add a dedicated subsection discussing the impact of GMM misspecification and report additional ablation experiments that vary the degree of noise heterogeneity to empirically characterize robustness. revision: partial

  2. Referee: [§3.2] §3.2 (GMM-weighted Calibration): The procedure for fitting the GMM and extracting instance-specific weights is described at a high level; it is unclear whether the number of components, initialization, or convergence criteria are chosen independently of the downstream recommendation objective. If these choices are tuned on the same noisy data used for BLTM estimation, the claimed consistency may be circular.

    Authors: We thank the referee for noting the need for greater procedural clarity. In the manuscript the GMM is fit to the distribution of per-instance losses using the Bayesian Information Criterion to select the number of components, k-means++ for initialization, and standard EM convergence thresholds. These choices are made once on the loss statistics before any BLTM estimation begins and are not subsequently tuned against the recommendation loss or validation metric. We will revise Section 3.2 to state this separation explicitly, add pseudocode for the full calibration pipeline, and include a short paragraph confirming that no downstream objective is used to select GMM hyperparameters. revision: yes

Circularity Check

0 steps flagged

No circularity: derivation claims rest on external theoretical analysis without self-referential reduction in provided text

full rationale

The abstract describes a GMM-weighted BLTM framework and asserts that theoretical analysis confirms consistency, full sample utilization, and variance reduction. However, no equations, derivation steps, or self-citations are exhibited in the supplied text that would allow inspection for self-definitional loops, fitted inputs renamed as predictions, or load-bearing self-citations. The GMM calibration is presented as a methodological choice whose correctness is supported by the claimed theory rather than by construction from the target result itself. Absent specific equations showing reduction to inputs (e.g., reliability scores defined directly from the same transition-matrix estimates they calibrate), the derivation chain cannot be shown to collapse. This is the normal case of a self-contained proposal whose validity hinges on external verification rather than internal circularity.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

The framework rests on the premise that implicit feedback noise can be modeled via a transition matrix whose bias can be corrected by instance-level GMM reliability scores; no explicit free parameters, axioms, or invented entities are detailed in the abstract.

pith-pipeline@v0.9.0 · 5730 in / 1127 out tokens · 28915 ms · 2026-05-21T05:45:21.740783+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

77 extracted references · 77 canonical work pages · 1 internal anchor

  1. [1]

    2008 Eighth IEEE international conference on data mining , pages=

    Collaborative filtering for implicit feedback datasets , author=. 2008 Eighth IEEE international conference on data mining , pages=. 2008 , organization=

    work page 2008

  2. [2]

    Proceedings of the 7th ACM international conference on Web search and data mining , pages=

    Modeling dwell time to predict click-level satisfaction , author=. Proceedings of the 7th ACM international conference on Web search and data mining , pages=

    work page

  3. [3]

    Proceedings of the 39th International ACM SIGIR conference on Research and Development in Information Retrieval , pages=

    Fast matrix factorization for online recommendation with implicit feedback , author=. Proceedings of the 39th International ACM SIGIR conference on Research and Development in Information Retrieval , pages=

    work page

  4. [4]

    Proceedings of the ninth ACM international conference on web search and data mining , pages=

    Collaborative denoising auto-encoders for top-n recommender systems , author=. Proceedings of the ninth ACM international conference on web search and data mining , pages=

    work page

  5. [5]

    Proceedings of the 10th ACM Conference on Recommender Systems , pages=

    Gaze prediction for recommender systems , author=. Proceedings of the 10th ACM Conference on Recommender Systems , pages=

    work page

  6. [6]

    Proceedings of the 40th International ACM SIGIR conference on Research and Development in Information Retrieval , pages=

    Neural factorization machines for sparse predictive analytics , author=. Proceedings of the 40th International ACM SIGIR conference on Research and Development in Information Retrieval , pages=

    work page

  7. [7]

    fBGD: Learning embeddings from positive unlabeled data with BGD , author=

    work page

  8. [8]

    Proceedings of the 12th ACM conference on recommender systems , pages=

    HOP-rec: high-order proximity for implicit recommendation , author=. Proceedings of the 12th ACM conference on recommender systems , pages=

    work page

  9. [9]

    Companion proceedings of the the web conference 2018 , pages=

    An improved sampler for bayesian personalized ranking by leveraging view data , author=. Companion proceedings of the the web conference 2018 , pages=

    work page 2018

  10. [10]

    The 41st international acm sigir conference on research & development in information retrieval , pages=

    Between clicks and satisfaction: Study on multi-phase user preferences and satisfaction for online news reading , author=. The 41st international acm sigir conference on research & development in information retrieval , pages=

    work page

  11. [11]

    Proceedings of the 42nd international ACM SIGIR conference on research and development in information retrieval , pages=

    An efficient adaptive transfer neural network for social-aware recommendation , author=. Proceedings of the 42nd international ACM SIGIR conference on research and development in information retrieval , pages=

    work page

  12. [12]

    Proceedings of the 25th ACM SIGKDD international conference on knowledge discovery & data mining , pages=

    Kgat: Knowledge graph attention network for recommendation , author=. Proceedings of the 25th ACM SIGKDD international conference on knowledge discovery & data mining , pages=

    work page

  13. [13]

    Proceedings of the 42nd international ACM SIGIR conference on Research and development in Information Retrieval , pages=

    Neural graph collaborative filtering , author=. Proceedings of the 42nd international ACM SIGIR conference on Research and development in Information Retrieval , pages=

    work page

  14. [14]

    Proceedings of the 42nd international ACM SIGIR conference on research and development in information retrieval , pages=

    To model or to intervene: A comparison of counterfactual and online learning to rank from user interactions , author=. Proceedings of the 42nd international ACM SIGIR conference on research and development in information retrieval , pages=

    work page

  15. [15]

    , author=

    Reinforced negative sampling for recommendation with exposure data. , author=. IJCAI , pages=. 2019 , organization=

    work page 2019

  16. [16]

    IEEE transactions on knowledge and data engineering , volume=

    Sampler design for bayesian personalized ranking by leveraging view data , author=. IEEE transactions on knowledge and data engineering , volume=. 2019 , publisher=

    work page 2019

  17. [17]

    International conference on machine learning , pages=

    Learning with bad training data via iterative trimmed loss minimization , author=. International conference on machine learning , pages=. 2019 , organization=

    work page 2019

  18. [18]

    Proceedings of the 42nd international ACM SIGIR conference on research and development in information retrieval , pages=

    Effects of user negative experience in mobile news streaming , author=. Proceedings of the 42nd international ACM SIGIR conference on research and development in information retrieval , pages=

    work page

  19. [19]

    Proceedings of the 43rd International ACM SIGIR conference on research and development in Information Retrieval , pages=

    Lightgcn: Simplifying and powering graph convolution network for recommendation , author=. Proceedings of the 43rd International ACM SIGIR conference on research and development in Information Retrieval , pages=

    work page

  20. [20]

    Proceedings of the 43rd international ACM SIGIR conference on research and development in information retrieval , pages=

    Sampler design for implicit feedback data by noisy-label robust learning , author=. Proceedings of the 43rd international ACM SIGIR conference on research and development in information retrieval , pages=

    work page

  21. [21]

    International Conference on Learning Representations , year=

    SELF: Learning to Filter Noisy Labels with Self-Ensembling , author=. International Conference on Learning Representations , year=

    work page

  22. [22]

    Proceedings of the web conference 2020 , pages=

    Reinforced negative sampling over knowledge graph for recommendation , author=. Proceedings of the web conference 2020 , pages=

    work page 2020

  23. [23]

    Proceedings of the 14th ACM international conference on web search and data mining , pages=

    Denoising implicit feedback for recommendation , author=. Proceedings of the 14th ACM international conference on web search and data mining , pages=

    work page

  24. [24]

    Proceedings of the 45th international ACM SIGIR conference on research and development in information retrieval , pages=

    Learning to denoise unreliable interactions for graph collaborative filtering , author=. Proceedings of the 45th international ACM SIGIR conference on research and development in information retrieval , pages=

    work page

  25. [25]

    Proceedings of the ACM web conference 2022 , pages=

    Learning robust recommenders through cross-model agreement , author=. Proceedings of the ACM web conference 2022 , pages=

    work page 2022

  26. [26]

    Proceedings of the 45th international ACM SIGIR conference on research and development in information retrieval , pages=

    Self-guided learning to denoise for robust recommendation , author=. Proceedings of the 45th international ACM SIGIR conference on research and development in information retrieval , pages=

    work page

  27. [27]

    Proceedings of the 31st ACM International Conference on Information & Knowledge Management , pages=

    LCD: Adaptive label correction for denoising music recommendation , author=. Proceedings of the 31st ACM International Conference on Information & Knowledge Management , pages=

    work page

  28. [28]

    Proceedings of the 29th ACM SIGKDD conference on knowledge discovery and data mining , pages=

    Efficient bi-level optimization for recommendation denoising , author=. Proceedings of the 29th ACM SIGKDD conference on knowledge discovery and data mining , pages=

    work page

  29. [29]

    CoRR , year=

    Automated Data Denoising for Recommendation , author=. CoRR , year=

    work page

  30. [30]

    ACM Transactions on Information Systems , volume=

    Towards robust neural graph collaborative filtering via structure denoising and embedding perturbation , author=. ACM Transactions on Information Systems , volume=. 2023 , publisher=

    work page 2023

  31. [31]

    Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining , pages=

    Double correction framework for denoising recommendation , author=. Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining , pages=

    work page

  32. [32]

    Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining , pages=

    Debiased recommendation with noisy feedback , author=. Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining , pages=

    work page

  33. [33]

    Proceedings of KDD cup 2011 , pages=

    Personalized ranking for non-uniformly sampled items , author=. Proceedings of KDD cup 2011 , pages=. 2012 , organization=

    work page 2011

  34. [34]

    arXiv preprint arXiv:2102.03482 , year=

    Understanding the interaction of adversarial training with noisy labels , author=. arXiv preprint arXiv:2102.03482 , year=

    work page arXiv

  35. [35]

    Advances in Neural Information Processing Systems , volume=

    Understanding and improving early stopping for learning with noisy labels , author=. Advances in Neural Information Processing Systems , volume=

    work page

  36. [36]

    Advances in Neural Information Processing Systems , volume=

    RSA: reducing semantic shift from aggressive augmentations for self-supervised learning , author=. Advances in Neural Information Processing Systems , volume=

    work page

  37. [37]

    International Conference on Learning Representations , year=

    DivideMix: Learning with Noisy Labels as Semi-supervised Learning , author=. International Conference on Learning Representations , year=

    work page

  38. [38]

    Machine learning , volume=

    Learning from noisy examples , author=. Machine learning , volume=. 1988 , publisher=

    work page 1988

  39. [39]

    IEEE Transactions on pattern analysis and machine intelligence , volume=

    Classification with noisy labels by importance reweighting , author=. IEEE Transactions on pattern analysis and machine intelligence , volume=. 2015 , publisher=

    work page 2015

  40. [40]

    International conference on machine learning , pages=

    Learning with bounded instance and label-dependent label noise , author=. International conference on machine learning , pages=. 2020 , organization=

    work page 2020

  41. [41]

    International conference on machine learning , pages=

    Confidence scores make instance-dependent label-noise learning possible , author=. International conference on machine learning , pages=. 2021 , organization=

    work page 2021

  42. [42]

    International Conference on Machine Learning , pages=

    Estimating instance-dependent bayes-label transition matrix using a deep neural network , author=. International Conference on Machine Learning , pages=. 2022 , organization=

    work page 2022

  43. [43]

    Advances in neural information processing systems , volume=

    Analysis of learning from positive and unlabeled data , author=. Advances in neural information processing systems , volume=

    work page

  44. [44]

    Advances in neural information processing systems , volume=

    Learning with noisy labels , author=. Advances in neural information processing systems , volume=

    work page

  45. [45]

    Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

    Making deep neural networks robust to label noise: A loss correction approach , author=. Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

    work page

  46. [46]

    International conference on machine learning , pages=

    Which is better for learning with noisy labels: the semi-supervised method or modeling label noise? , author=. International conference on machine learning , pages=. 2023 , organization=

    work page 2023

  47. [47]

    Advances in neural information processing systems , volume=

    Dual t: Reducing estimation error for transition matrix in label-noise learning , author=. Advances in neural information processing systems , volume=

    work page

  48. [48]

    International conference on machine learning , pages=

    Provably end-to-end label-noise learning without anchor points , author=. International conference on machine learning , pages=. 2021 , organization=

    work page 2021

  49. [49]

    Advances in Neural Information Processing Systems , volume=

    Class-dependent label-noise learning with cycle-consistency regularization , author=. Advances in Neural Information Processing Systems , volume=

    work page

  50. [50]

    Advances in neural information processing systems , volume=

    Part-dependent label noise: Towards instance-dependent label noise , author=. Advances in neural information processing systems , volume=

    work page

  51. [51]

    IEEE Transactions on Pattern Analysis and Machine Intelligence , year=

    Improving the Instance-Dependent Transition Matrix Estimation by Exploiting Self-Supervised Learning , author=. IEEE Transactions on Pattern Analysis and Machine Intelligence , year=

    work page

  52. [52]

    Advances in Neural Information Processing Systems , volume=

    Instance-dependent label-noise learning under a structural causal model , author=. Advances in Neural Information Processing Systems , volume=

    work page

  53. [53]

    International Conference on Learning Representations , year=

    Learning with Noisy Labels Revisited: A Study Using Real-World Human Annotations , author=. International Conference on Learning Representations , year=

    work page

  54. [54]

    International conference on learning representations , year=

    Robust early-learning: Hindering the memorization of noisy labels , author=. International conference on learning representations , year=

    work page

  55. [55]

    ACM Transactions on Intelligent Systems and Technology , volume=

    Robust recommender systems with rating flip noise , author=. ACM Transactions on Intelligent Systems and Technology , volume=. 2024 , publisher=

    work page 2024

  56. [56]

    Neural Networks , volume=

    Adaptive estimation of instance-dependent noise transition matrix for learning with instance-dependent label noise , author=. Neural Networks , volume=. 2025 , publisher=

    work page 2025

  57. [57]

    IEEE Access , year=

    Joint Modeling of Prediction and Behavioral Patterns for Reliable Recommendation with Implicit Feedback , author=. IEEE Access , year=

    work page

  58. [58]

    Machine Learning , volume=

    Learning from binary labels with instance-dependent noise , author=. Machine Learning , volume=. 2018 , publisher=

    work page 2018

  59. [59]

    IEEE Transactions on Pattern Analysis and Machine Intelligence , year=

    Practically Unbiased Pairwise Loss for Recommendation With Implicit Feedback , author=. IEEE Transactions on Pattern Analysis and Machine Intelligence , year=

    work page

  60. [60]

    BPR: Bayesian Personalized Ranking from Implicit Feedback

    BPR: Bayesian personalized ranking from implicit feedback , author=. arXiv preprint arXiv:1205.2618 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  61. [61]

    Computer , volume=

    Matrix factorization techniques for recommender systems , author=. Computer , volume=. 2009 , publisher=

    work page 2009

  62. [62]

    Proceedings of the 47th International ACM SIGIR Conference on Research and Development in Information Retrieval , pages=

    Denoising diffusion recommender model , author=. Proceedings of the 47th International ACM SIGIR Conference on Research and Development in Information Retrieval , pages=

    work page

  63. [63]

    Proceedings of the 46th international ACM SIGIR conference on research and development in information retrieval , pages=

    Diffusion recommender model , author=. Proceedings of the 46th international ACM SIGIR conference on research and development in information retrieval , pages=

    work page

  64. [64]

    IEEE Transactions on Neural Networks and Learning Systems , year=

    Cognition-driven structural prior for instance-dependent label transition matrix estimation , author=. IEEE Transactions on Neural Networks and Learning Systems , year=

    work page

  65. [65]

    IEEE Transactions on Pattern Analysis and Machine Intelligence , volume=

    A parametrical model for instance-dependent label noise , author=. IEEE Transactions on Pattern Analysis and Machine Intelligence , volume=. 2023 , publisher=

    work page 2023

  66. [66]

    Advances in neural information processing systems , volume=

    Are anchor points really indispensable in label-noise learning? , author=. Advances in neural information processing systems , volume=

    work page

  67. [67]

    ACM Computing Surveys , volume=

    Multimodal recommender systems: A survey , author=. ACM Computing Surveys , volume=. 2024 , publisher=

    work page 2024

  68. [68]

    Proceedings of the 46th international ACM SIGIR conference on research and development in information retrieval , pages=

    Causal recommendation: Progresses and future directions , author=. Proceedings of the 46th international ACM SIGIR conference on research and development in information retrieval , pages=

    work page

  69. [69]

    IEEE Computational Intelligence Magazine , volume=

    Recent developments in recommender systems: A survey , author=. IEEE Computational Intelligence Magazine , volume=. 2024 , publisher=

    work page 2024

  70. [70]

    Data Science and Engineering , pages=

    On-device recommender systems: A comprehensive survey , author=. Data Science and Engineering , pages=. 2025 , publisher=

    work page 2025

  71. [71]

    IEEE Transactions on Big Data , year=

    Unified representation learning for discrete attribute enhanced completely cold-start recommendation , author=. IEEE Transactions on Big Data , year=

    work page

  72. [72]

    ACM Transactions on Intelligent Systems and Technology , volume=

    Mitigating recommendation biases via group-alignment and global-uniformity in representation learning , author=. ACM Transactions on Intelligent Systems and Technology , volume=. 2024 , publisher=

    work page 2024

  73. [73]

    Companion Proceedings of the ACM Web Conference 2024 , pages=

    Collaborative-enhanced prediction of spending on newly downloaded mobile games under consumption uncertainty , author=. Companion Proceedings of the ACM Web Conference 2024 , pages=

    work page 2024

  74. [74]

    Information Processing & Management , volume=

    Matrix factorization recommender based on adaptive Gaussian differential privacy for implicit feedback , author=. Information Processing & Management , volume=. 2024 , publisher=

    work page 2024

  75. [75]

    Proceedings of the ACM Web Conference 2024 , pages=

    Intersectional two-sided fairness in recommendation , author=. Proceedings of the ACM Web Conference 2024 , pages=

    work page 2024

  76. [76]

    ACM Transactions on Recommender Systems , volume=

    Breaking feedback loops in recommender systems with causal inference , author=. ACM Transactions on Recommender Systems , volume=. 2025 , publisher=

    work page 2025

  77. [77]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    Instance-dependent label-noise learning with manifold-regularized transition matrix estimation , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    work page