Beyond Self-Consistency: Loss-Balanced Perturbation-Based Regularization Improves Industrial-Scale Ads Ranking

Bo Long; Hamid Eghbalzadeh; Ilqar Ramazanli; Jiaxiang Fu; Kaushik Rangadurai; Sem Park; Xiaoyi Liu; Xue Feng; Yang Wang

REVIEW 4 major objections 5 minor 15 references

Beyond Self-Consistency: Loss-Balanced Perturbation-Based Regularization Improves Industrial-Scale Ads Ranking

T0 review · 4 major / 5 minor · reviewed 2026-08-09 · deepseek-v4-flash

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

Pith's one-line read Perturbation-based regularization, applied for the first time at billion-user scale, improves industrial ads ranking.

desk verdict LSPR is a minor twist on training-with-noise with a plausible industrial story, but the headline comparison against SCR is confounded and the evidence is too thin to accept as-is. read the letter →

arxiv 2502.18478 v1 pith:MHC3V7KG submitted 2025-02-05 cs.IR cs.AIcs.LG

Ilqar Ramazanli , Hamid Eghbalzadeh , Xiaoyi Liu , Yang Wang , Jiaxiang Fu , Kaushik Rangadurai , Sem Park , Bo Long

show 1 more author

Xue Feng

This is my paper · ORCID

classification cs.IRcs.AIcs.LG

keywords adsrankingperturbation-basedregularizationself-consistencyloss-balancedsmallperturbationnormalizedentropyindustrialrecommendersystemsdataaugmentationclick-throughrateprediction

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

The paper is trying to establish that perturbation-based regularization—adding small noisy copies of training examples—can work at the scale of a billion-user ads ranking system, and that a specific new variant, Loss-Balanced Small Perturbation Regularization (LSPR), beats the established Self-Consistency Regularization (SCR) baseline. LSPR adds perturbed examples to the training batch but down-weights their supervised loss by a factor $\lambda < 1$, instead of adding an auxiliary loss that pushes perturbed and original predictions closer together. The paper reports relative Normalized Entropy gains of about 0.1% to 0.3% offline with SCR, and consistent 0.10% to 0.13% relative NE gains for LSPR over the production baseline in the final-stage ranker, with LSPR outperforming SCR at every data-availability level. It also reports 0.1% to 0.2% top-line improvements in online launches. A sympathetic reader would care because the method is simple enough to deploy across many production deep learning models and is claimed to generalize across ranking stages, surfaces, and traffic.

What carries the argument

The load-bearing object is the LSPR loss, a weighted sum of the ordinary supervised loss on clean examples and the same supervised loss on perturbed copies: $L_{\mathrm{LSPR}}(y, \hat{y}, \hat{y}') = L_{\mathrm{supervised}}(y, \hat{y}) + \lambda L_{\mathrm{supervised}}(y, \hat{y}')$, with $\lambda < 1$. Perturbations are drawn from the same distribution class as the features—Gaussian noise for dense features and dropout for sparse features—so the noisy copy keeps the original label. The mechanism differs from SCR in that LSPR never tries to make clean and perturbed predictions equal; it treats the perturbed point as a down-weighted supervised example, which the paper argues disturbs learning dynamics less and doubles the effective batch size. This loss is what carries both the offline NE gains and the claimed online improvements, and the paper's linear-model analysis uses it to show better alignment with the optimal weight $W^*$ than SCR.

What would settle it

Re-run the final-stage ranker experiment from Table 2 multiple times with different random seeds and data subsamples, computing the mean and confidence interval of the relative NE gain for LSPR versus SCR at each data-availability level; if LSPR's gain is not consistently positive across repetitions, the paper's central claim that LSPR performs consistently better than SCR collapses.

Watch

Extended reading notes

Core claim

The central claim is that a regularizer which simply copies each training example, adds feature-consistent noise (Gaussian perturbation to dense features and dropout to sparse features), and includes the noisy copy in the supervised loss with a scalar weight $\lambda < 1$ improves a billion-scale ads ranking system. The paper's stated discovery is that this LSPR loss, $L_{\mathrm{LSPR}} = L_{\mathrm{supervised}}(y, \hat{y}) + \lambda L_{\mathrm{supervised}}(y, \hat{y}')$, performs consistently better than Self-Consistency Regularization, which instead enforces consistency between the predictions on original and perturbed inputs through an MSE auxiliary term. In the final-stage ranker experiments, LSPR yields 0.13%, 0.11%, and 0.10% relative NE gains over the production baseline at 33%, 66%, and 100% data availability, versus SCR's 0.10%, 0.08%, and 0.07%; the paper also reports that the offline gains carry over to online top-line metrics at 0.1% to 0.2%. The numerical analysis with linear models is offered as evidence that LSPR's gradient updates align better with the optimal model weights and reach lower weight-space error than SCR.

Load-bearing premise

The load-bearing premise is that the reported relative Normalized Entropy differences of 0.07% to 0.13% over the production baseline are real and not run-to-run noise; the paper gives no confidence intervals, repeated runs, or significance tests for these differences.

Editorial extensions

If this is right

LSPR can be dropped into any supervised deep learning model by cloning each batch, perturbing features, and down-weighting the clone's loss; no auxiliary head or consistency loss is needed.
At 33%, 66%, and 100% of training data, LSPR's relative NE gain over the production baseline is consistently larger than SCR's, so the advantage holds as data becomes more available.
Because retrieval-stage and early-stage ranking models also improved under perturbation-based regularization, the technique applies across the full multi-stage ranking pipeline, not just the final ranker.
If the online 0.1% to 0.2% top-line improvements are stable across multiple launches, the method is usable in continuously trained production environments, not only offline retraining.
The improvement is measured relative to production baselines that were already state-of-the-art, so the resulting ranking quality gain is on top of mature industrial ads systems.

Reading between the lines

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

Editorial inference: the paper's numerical analysis only covers linear models, so an immediate extension is to test whether LSPR's better alignment with optimal weights on linear systems also holds for the deep final-stage ranker.
Editorial inference: because LSPR only changes how perturbed copies are weighted, it could be combined with other augmentation families such as feature masking or mixup-style interpolation; the paper only tests Gaussian noise and dropout.
Editorial inference: the reported gains are small (0.07% to 0.13% relative NE), and without confidence intervals or repeated-run statistics, a reader should treat the offline advantage as a point estimate rather than a proven difference; this is a gap the paper does not address.
Editorial inference: the paper mentions that perturbation-dependent weights are a future direction; a natural testable extension is to weight perturbed copies by their noise magnitude or by feature importance rather than a uniform $\lambda$.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note

read the letter

Colleague,

The new piece here is LSPR, which is simply training with noisy copies of examples, downweighted by lambda in the loss, applied at billion-scale in Meta's ads ranking. That is genuinely new as a deployment claim; the algorithm itself is a weighted version of Bishop's training-with-noise, which they cite. The paper is worth a look because it shows perturbation-based regularization can survive industrial-scale sparse features and production constraints.

What it does well: the method is simple to state, the system description gives some idea of where such regularization helps across retrieval, early, and final ranking, and the offline gains of 0.1-0.3% relative NE are consistent in direction with the online claim of 0.1-0.2%. The authors also cite Bishop and acknowledge the lineage, which is more than some industrial papers do.

The soft spots are real. The main comparison in Table 2 is not controlled: LSPR perturbs the entire batch and doubles batch size, SCR perturbs only a small fraction, so the 0.03 percentage point gap between LSPR and SCR could just be more augmentation per step. The paper reports no confidence intervals, no repeated runs, and no significance tests for these differences. Hyperparameters are explicitly deferred to "the final version," so the setup cannot be reproduced from the text. The online results are given without launch statistics. The numerical analysis in Section 5.1 is a simulation with chosen lambda and omega, not a derivation, and it uses MSE on a linear model while the real experiments use NE on a deep network, so it is illustrative at best. The "first of its kind" claim is also overstated given that noise injection has a long history; what is first is the specific deployment, not the idea.

None of this proves the method does not work. The central claim, that a downweighted noise term helps at scale, is plausible and the offline direction matches the online direction. But the evidence as written does not separate LSPR's loss-balancing mechanism from the extra perturbed data it consumes.

Give this paper a serious referee. It reports a real deployed change to a billion-scale system, and the confound in Table 2 is exactly what a good reviewer should force the authors to untangle. I would not cite it in its current form because the method is not novel enough and the results are not well enough supported. I would bring it to a reading group, though, as a case study in how industrial deployment papers can under-report comparison setup.

Referee Report

4 major / 5 minor

Summary. The paper proposes Loss-Balanced Small Perturbation Regularization (LSPR), a perturbation-based regularization method that adds noisy copies of training examples to the training batch and down-weights them in the supervised loss, in contrast to Self-Consistency Regularization (SCR), which adds an auxiliary MSE consistency loss. The authors report offline relative Normalized Entropy (NE) gains of 0.1%–0.3% for SCR and 0.07%–0.13% for LSPR over a production baseline in a multi-stage industrial ads ranking system, and claim an online top-line improvement of 0.1%–0.2% from LSPR. They also present a numerical analysis on a small linear model (Section 5.1) suggesting that LSPR aligns better with the optimal weights than SCR. The paper's central claim is that LSPR consistently outperforms SCR across data-availability settings and is the first perturbation-based regularization deployed in a billion-scale ads ranking system.

Significance. If the empirical claims are correct, the contribution is practically significant: a simple, architecture-agnostic regularizer that yields 0.1%–0.2% relative NE gains at billion-user scale is valuable for industrial recommender systems, and the paper provides a clear algorithmic description (Algorithm 1, Eq. 2) that others could implement. The industrial deployment description and the multi-stage integration details are useful to practitioners. However, the evidence presented is currently too weak to establish the central superiority claim: the headline comparison in Table 2 is an uncontrolled experiment, no uncertainty quantification is provided for the small offline differences, and the numerical analysis in Section 5.1 is a synthetic simulation rather than a derivation or a production-scale evaluation. The claimed novelty of being the 'first' industrial-scale deployment is also sensitive to the exact scope of related work (e.g., Yao et al., 2021, report large-scale item recommendations with perturbation-based SSL).

major comments (4)

[Section 5.2.2, Table 2] The headline comparison between LSPR and SCR is uncontrolled. The text states that for LSPR 'we perturbed the entire batch each time, leading to doubled batch size,' whereas for SCR 'we only perturbed a small fraction of points in each batch.' Thus Table 2 simultaneously varies the loss-balancing mechanism and the augmentation budget/effective batch size. The 0.02–0.03 percentage-point gaps between LSPR and SCR (0.13 vs 0.10, 0.11 vs 0.08, 0.10 vs 0.07) could plausibly be explained by the larger number of perturbed examples seen by LSPR, rather than by the loss-balancing rule itself. To support the claim that LSPR 'performs consistently better compared to SCR,' the authors should report a controlled comparison with matched perturbation budgets (e.g., equal numbers of perturbed examples per step or equal total compute), or add ablations that isolate the loss-balancing component.
[Table 2 and Section 5.2.3] The offline gains and the online claim lack uncertainty quantification. The relative NE differences in Table 2 are 0.07%–0.13% relative to baseline, which are small enough to be within run-to-run noise for industrial training pipelines, yet the paper reports no confidence intervals, no repeated runs, and no significance tests. Likewise, the online claim of '0.1% to 0.2% relative improvement' is described only qualitatively ('consistently across multiple launches') with no launch statistics, segment-level variance, or guardrail metrics. Without these, the central empirical claim that LSPR is consistently better than SCR is not statistically supported.
[Section 5.1, Eqs. (12)–(15)] The numerical analysis does not provide a derivation of LSPR's advantage; it writes down the LSPR and SCR gradient updates as definitions and then simulates a two-layer linear model with MSE loss and selected hyperparameters (omega in {0.1, 0.9}, lambda in {0.001, 1}, eta = 1.4). Because the simulation uses a different loss (MSE) and architecture than the production setting (NE on a large ranking model), and because the hyperparameters are chosen by the authors, Figure 4 does not independently establish that LSPR outperforms SCR in the real system. The authors should either provide an analytic characterization of when LSPR dominates SCR (e.g., in terms of the bias-variance trade-off of the gradient updates) or base the superiority claim on controlled real-data experiments with proper ablations.
[Section 4.4] Reproducibility is severely limited by missing hyperparameter details. Section 4.4 states that the dense-feature perturbation distribution, sparse-feature dropout rate, and loss weight values 'will be added to the final version of our paper.' Without these values, and without any description of the search ranges or the final chosen settings, the experiments in Table 2 cannot be reconstructed. Since the central claim is empirical and the gains are small, the exact hyperparameter choices are load-bearing and must be reported in the main text or an appendix.

minor comments (5)

[Abstract, Section 1] There are grammatical errors that should be corrected, e.g., 'We have successfully demonstrate' in the abstract and 'weight them down in the the loss term calculation' in Section 1.
[Section 5.1, Eqs. (14)–(15)] The text refers to 'SCP weight updates' when the intended acronym is SCR; this typo appears twice and should be fixed.
[Section 4.2, Eq. (2)] The sentence defining the LSPR loss, 'where ŷ′_i is the model's prediction on the perturbed input x′_i traditional supervised loss function,' is incomplete and should be rewritten.
[Figure 4 caption] The caption says 'ω denotes noise sample weight,' but in Eqs. (12)–(15) ω appears as the perturbation magnitude scaling the noise z; the notation and terminology should be aligned.
[References] Several references are malformed or incomplete, e.g., 'Wang F, Wang Y, Li D, Gu H, Lu T, Zhang P, and Gu N' and 'LeCun Y, Bengio Y, and Hinton G'; these should be formatted consistently with the rest of the bibliography.

Circularity Check

0 steps flagged · score 2.0 of 10

No material circularity found: the core claims are empirical measurements against production baselines, and no load-bearing derivation reduces to its own inputs.

full rationale

The central claims that LSPR improves relative Normalized Entropy and outperforms SCR are empirical. LSPR is defined by Eq. (2), and the reported gains in Tables 1-2 and the online experiments are measured outcomes, not consequences of that definition. The numerical analysis in Section 5.1 simulates the two gradient-update rules and observes alignment with W*; although the setup is author-chosen, this is an experimental illustration, not a case where Eqs. (12)-(15) are assumed to imply the conclusion. The unmatched augmentation budgets in Section 5.2.2, where LSPR perturbs the whole batch and doubles batch size while SCR perturbs a small fraction, are a real confound for the LSPR-versus-SCR comparison, but confounding is not circularity. Section 4.4's omission of hyperparameter values is a reproducibility limitation, not a circular step. The only author self-citation (Eghbalzadeh et al. 2024) appears in a general data-augmentation discussion in Section 2.4 and is not load-bearing. Therefore no circular step can be exhibited, and the score remains in the low, non-significant range.

Assumptions & free parameters 4 free parameters · 6 assumptions · 0 invented entities

The central claims rest on a small set of tuning parameters (loss weight, noise scale, dropout rate) that are not fully disclosed, plus assumptions that perturbed examples keep their labels, that Gaussian/dropout perturbations are appropriate for ads features, and that a 2-layer linear simulation captures real ranking dynamics. No new entities are introduced.

free parameters (4)

loss weight lambda = not disclosed
Balances the supervised loss on original vs perturbed data in Eq. 2; the paper says it is tuned by search but omits final values.
perturbation magnitude omega = 0.1 or 0.9 in toy analysis; not disclosed for production
Controls noise strength for Gaussian perturbation; the toy experiments use omega = {0.1, 0.9}, and production values are not given.
noise standard deviation sigma = not disclosed
Scale of Gaussian noise for dense features, referenced in Eqs. 12-13 but no production value is provided.
sparse feature dropout rate = not disclosed
A small dropout rate applied to sparse features; the exact rate is not provided.

assumptions (6)

domain assumption Perturbed examples retain the label of the original example (label-preserving perturbation).
Used in Eq. 2 where L_supervised(y, yhat') is computed with the original label y; if noise crosses decision boundaries frequently, this assumption breaks.
domain assumption Noise distribution matched to feature type (Gaussian for dense, dropout for sparse) is a valid augmentation for ads ranking.
Section 4.1 and 4.4 assume these perturbations preserve contextual information; no validation is provided.
domain assumption A 2-layer linear model with MSE loss captures the relevant learning dynamics of industrial deep ranking models.
Section 5.1.1 restricts the analysis to linear networks and synthetic data, then uses the results to draw conclusions about LSPR vs SCR without proof of transfer.
domain assumption Normalized Entropy (NE) gains of about 0.1% are statistically meaningful for the production system.
Tables 1 and 2 report relative NE gains without variance or significance testing; the paper treats them as reliable.
domain assumption Production baseline models are appropriate state-of-the-art comparators.
Section 5.3 says baselines are production models but gives no architecture or metric details, making the comparison hard to audit.
standard math Standard stochastic gradient descent and MSE error as formulated in Section 5.1.1.
The gradient updates in Eqs. 12-15 are algebraic consequences of the defined losses; this background is standard.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Beyond Self-Consistency: Loss-Balanced Perturbation-Based Regularization Improves Industrial-Scale Ads Ranking." pith.science (2026). https://pith.science/paper/MHC3V7KG

@misc{pith2026250218478,
  author       = {Pith},
  title        = {Pith review of: Beyond Self-Consistency: Loss-Balanced Perturbation-Based Regularization Improves Industrial-Scale Ads Ranking},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MHC3V7KG}},
  note         = {Machine review of arXiv:2502.18478}
}

read the original abstract

Perturbation-based regularization techniques address many challenges in industrial-scale large models, particularly with sparse labels, and emphasize consistency and invariance for perturbation in model predictions. One of the popular regularization techniques has been various forms of self-consistency, which involve making small modifications to input data while preserving contextual information and enforcing similar predictions through auxiliary loss functions. In this work, we explore the first successful application of perturbation-based regularization algorithms in large-scale ads ranking models, and further propose a novel regularization algorithm, namely, Loss-Balanced Small Perturbation Regularization (LSPR) that can be used in potentially any deep learning model. We have successfully demonstrate that both Self-Consistency Regularization approaches (SCR) and LSPR are scalable and can improve ads delivery systems. By conducting industrial-scale experiments, and numerical analysis, we additionally show that our proposed LSPR, performs consistently better compared to SCR, across various groups and signal availability setups. Finally, we report a successful application of the proposed LSPR in a billion-scale industrial ranking system, which to the best of our knowledge, is the first of its kind, and it is specially designed to address the various scalability challenges (e.g, various surfaces, geological locations, clients and so on) as we will mention in this paper.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

15 extracted references · 9 canonical work pages

[1]

On the factory floor: Ml engineering for industrial-scale ads recommendation models

Rohan Anil, Sandra Gadanho, Da Huang, Nijith Jacob, Zhuoshu Li, Dong Lin, Todd Phillips, Cristina Pop, Kevin Regan, Gil I Shamir, et al. On the factory floor: Ml engineering for industrial-scale ads recommendation models. arXiv preprint arXiv:2209.05310,

arXiv
[9]

Explicit regularization in overparametrized models via noise injection

Antonio Orvieto, Anant Raj, Hans Kersting, and Francis Bach. Explicit regularization in overparametrized models via noise injection. In Francisco Ruiz, Jennifer Dy, and Jan-Willem van de Meent, editors,Proceedings of The 26th International Conference on Artificial Intelligence and Statistics, volume 206 ofProceedings of Machine Learning Research, pages 72...

work page 2023
[11]

Feature Dropout: Revisiting the Role of Augmentations in Contrastive Learning

Nitish Srivastava, Geoffrey Hinton, Alex Krizhevsky, Ilya Sutskever, and Ruslan Salakhutdinov. Dropout: A simple way to prevent neural networks from overfitting.Journal of Machine Learning Research, 15(56):1929–1958, 2014a. http://jmlr.org/papers/v15/srivastava14a.html. Nitish Srivastava, Geoffrey Hinton, Alex Krizhevsky, Ilya Sutskever, and Ruslan Salakh...

work page Pith review arXiv 1929
[14]

Dhen: A deep and hierarchical ensemble network for large-scale click-through rate prediction

Buyun Zhang, Liang Luo, Xi Liu, Jay Li, Zeliang Chen, Weilin Zhang, Xiaohan Wei, Yuchen Hao, Michael Tsang, Wenjun Wang, Yang Liu, Huayu Li, Yasmine Badr, Jongsoo Park, Jiyan Yang, Dheevatsa Mudigere, and Ellie Wen. Dhen: A deep and hierarchical ensemble network for large-scale click-through rate prediction. In4th Workshop on Deep Learning Practice and Th...

work page 2022
[15]

mixup: Beyond empirical risk minimization

Hongyi Zhang, Moustapha Cisse, Yann N Dauphin, and David Lopez-Paz. mixup: Beyond empirical risk minimization. arXiv preprint arXiv:1710.09412,

arXiv
[26]

Hao Wang, Naiyan Wang, and Dit-Yan Yeung

Curran Associates, Inc., 2013.https://proceedings.neurips.cc/paper_files/paper/2013/file/ 38db3aed920cf82ab059bfccbd02be6a-Paper.pdf. Hao Wang, Naiyan Wang, and Dit-Yan Yeung. Collaborative deep learning for recommender systems. pages 1235–1244,

work page 2013
[2003]

Deep learning.nature, pages 436–44., 2015 May

LeCun Y, Bengio Y, and Hinton G. Deep learning.nature, pages 436–44., 2015 May. Tiansheng Yao, Xinyang Yi, Derek Zhiyuan Cheng, Felix Yu, Ting Chen, Aditya Menon, Lichan Hong, Ed H Chi, Steve Tjoa, Jieqi Kang, et al. Self-supervised learning for large-scale item recommendations. InProceedings of the 30th ACM International Conference on Information & Knowl...

work page 2015
[2007]

Samarth Sinha and Adji Bousso Dieng

doi: 10.1186/s40537-019-0197-0.https://journalofbigdata.springeropen.com/counter/pdf/ 10.1186/s40537-019-0197-0.pdf . Samarth Sinha and Adji Bousso Dieng. Consistency regularization for variational auto-encoders.Advances in Neural Information Processing Systems, 34:12943–12954,

work page doi:10.1186/s40537-019-0197-0.https://journalofbigdata.springeropen.com/counter/pdf/

Show all 15 references

[2014]

Noise stability regularization for improving bert fine-tuning

Hang Hua, Xingjian Li, Dejing Dou, Chengzhong Xu, and Jiebo Luo. Noise stability regularization for improving bert fine-tuning. In Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pa...

2021
[2016]

Devlin, M.-W

J. Devlin, M.-W. Chang, K. Lee, , and K. Toutanova. Bert: Pretraining of deep bidirectional transformers for language understanding. arXiv preprint arXiv:1810.04805,

arXiv
[2019]

Simcse: Simple contrastive learning of sentence embeddings.arXiv preprint arXiv:2104.08821,

Tianyu Gao, Xingcheng Yao, and Danqi Chen. Simcse: Simple contrastive learning of sentence embeddings.arXiv preprint arXiv:2104.08821,

arXiv
[2020]

Diffcse: Difference-based contrastive learning for sentence embeddings

Yung-Sung Chuang, Rumen Dangovski, Hongyin Luo, Yang Zhang, Shiyu Chang, Marin Soljačić, Shang-Wen Li, Wen-tau Yih, Yoon Kim, and James Glass. Diffcse: Difference-based contrastive learning for sentence embeddings. arXiv preprint arXiv:2204.10298,

arXiv
[2021]

Selfmatch: Combining contrastive self-supervision and consistency for semi-supervised learning.arXiv preprint arXiv:2101.06480,

Byoungjip Kim, Jinho Choo, Yeong-Dae Kwon, Seongho Joe, Seungjai Min, and Youngjune Gwon. Selfmatch: Combining contrastive self-supervision and consistency for semi-supervised learning.arXiv preprint arXiv:2101.06480,

arXiv
[2022]

Deep learning recommendation model for personalization and recommendation systems.arXiv preprint arXiv:1906.00091,

Maxim Naumov, Dheevatsa Mudigere, Hao-Jun Michael Shi, Jianyu Huang, Narayanan Sundaraman, Jongsoo Park, Xiaodong Wang, Udit Gupta, Carole-Jean Wu, Alisson G Azzolini, et al. Deep learning recommendation model for personalization and recommendation systems.arXiv preprint arXiv...

1906 arXiv
[2024]

doi: https://doi.org/10.1016/j.ins.2023.119838

ISSN 0020-0255. doi: https://doi.org/10.1016/j.ins.2023.119838. Erik Englesson and Hossein Azizpour. Consistency regularization can improve robustness to label noise.arXiv preprint arXiv:2110.01242,

2023

Pith tools

Reviewed August 9, 2026 · model on record in the stance chip above.

[1] [1]

On the factory floor: Ml engineering for industrial-scale ads recommendation models

Rohan Anil, Sandra Gadanho, Da Huang, Nijith Jacob, Zhuoshu Li, Dong Lin, Todd Phillips, Cristina Pop, Kevin Regan, Gil I Shamir, et al. On the factory floor: Ml engineering for industrial-scale ads recommendation models. arXiv preprint arXiv:2209.05310,

arXiv

[2] [9]

Explicit regularization in overparametrized models via noise injection

Antonio Orvieto, Anant Raj, Hans Kersting, and Francis Bach. Explicit regularization in overparametrized models via noise injection. In Francisco Ruiz, Jennifer Dy, and Jan-Willem van de Meent, editors,Proceedings of The 26th International Conference on Artificial Intelligence and Statistics, volume 206 ofProceedings of Machine Learning Research, pages 72...

work page 2023

[3] [11]

Feature Dropout: Revisiting the Role of Augmentations in Contrastive Learning

Nitish Srivastava, Geoffrey Hinton, Alex Krizhevsky, Ilya Sutskever, and Ruslan Salakhutdinov. Dropout: A simple way to prevent neural networks from overfitting.Journal of Machine Learning Research, 15(56):1929–1958, 2014a. http://jmlr.org/papers/v15/srivastava14a.html. Nitish Srivastava, Geoffrey Hinton, Alex Krizhevsky, Ilya Sutskever, and Ruslan Salakh...

work page Pith review arXiv 1929

[4] [14]

Dhen: A deep and hierarchical ensemble network for large-scale click-through rate prediction

Buyun Zhang, Liang Luo, Xi Liu, Jay Li, Zeliang Chen, Weilin Zhang, Xiaohan Wei, Yuchen Hao, Michael Tsang, Wenjun Wang, Yang Liu, Huayu Li, Yasmine Badr, Jongsoo Park, Jiyan Yang, Dheevatsa Mudigere, and Ellie Wen. Dhen: A deep and hierarchical ensemble network for large-scale click-through rate prediction. In4th Workshop on Deep Learning Practice and Th...

work page 2022

[5] [15]

mixup: Beyond empirical risk minimization

Hongyi Zhang, Moustapha Cisse, Yann N Dauphin, and David Lopez-Paz. mixup: Beyond empirical risk minimization. arXiv preprint arXiv:1710.09412,

arXiv

[6] [26]

Hao Wang, Naiyan Wang, and Dit-Yan Yeung

Curran Associates, Inc., 2013.https://proceedings.neurips.cc/paper_files/paper/2013/file/ 38db3aed920cf82ab059bfccbd02be6a-Paper.pdf. Hao Wang, Naiyan Wang, and Dit-Yan Yeung. Collaborative deep learning for recommender systems. pages 1235–1244,

work page 2013

[7] [2003]

Deep learning.nature, pages 436–44., 2015 May

LeCun Y, Bengio Y, and Hinton G. Deep learning.nature, pages 436–44., 2015 May. Tiansheng Yao, Xinyang Yi, Derek Zhiyuan Cheng, Felix Yu, Ting Chen, Aditya Menon, Lichan Hong, Ed H Chi, Steve Tjoa, Jieqi Kang, et al. Self-supervised learning for large-scale item recommendations. InProceedings of the 30th ACM International Conference on Information & Knowl...

work page 2015

[8] [2007]

Samarth Sinha and Adji Bousso Dieng

doi: 10.1186/s40537-019-0197-0.https://journalofbigdata.springeropen.com/counter/pdf/ 10.1186/s40537-019-0197-0.pdf . Samarth Sinha and Adji Bousso Dieng. Consistency regularization for variational auto-encoders.Advances in Neural Information Processing Systems, 34:12943–12954,

work page doi:10.1186/s40537-019-0197-0.https://journalofbigdata.springeropen.com/counter/pdf/