pith. sign in

arxiv: 2511.10161 · v2 · submitted 2025-11-13 · 💻 cs.AI · cs.LG

DenoGrad: A Gradient-Based Framework for Data Refinement in Tabular and Time-Series Learning

J. Javier Alonso-Ramos , Ignacio Aguilera-Martos , Francisco Herrera , Andr\'es Herrera-Poyatos This is my paper

Pith reviewed 2026-05-17 22:43 UTC · model grok-4.3

classification 💻 cs.AI cs.LG
keywords data refinementgradient optimizationdenoisingtabular learningtime series forecastingdata-centric AIneural network guided refinement
0
0 comments X

The pith

DenoGrad refines noisy data by optimizing inputs via gradients from a fixed pretrained neural network, improving predictions on tabular and time-series tasks.

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

The paper proposes DenoGrad to address limitations in current denoising methods that rely on strict assumptions or clean reference data. It uses a pretrained model kept fixed while adjusting the input data through gradient-based optimization to correct noise. This is applied to tabular regression and time-series forecasting with a consensus strategy for temporal consistency. Experiments across ten real-world datasets demonstrate better predictive performance and maintained statistical properties of the data. The method also shows benefits for clean datasets by providing regularization effects.

Core claim

DenoGrad is a gradient-based framework for data refinement that leverages a pretrained neural network to iteratively correct noisy observations by optimizing the input space while keeping the model fixed, applicable to tabular regression and time-series forecasting with a consensus-based strategy for temporally coherent updates, yielding consistent improvements in downstream predictive performance while preserving the statistical structure on ten real-world datasets and improving generalization in clean data.

What carries the argument

Gradient-based optimization of the input data using a fixed pretrained neural network to reduce prediction loss.

If this is right

  • Refined data leads to better predictive performance in downstream tasks.
  • Statistical structure including distributions and correlations remains largely unchanged.
  • Nominally clean datasets can see improved generalization as a regularization effect.
  • The approach applies to both tabular and sequential data with appropriate adjustments for coherence.
  • Model-guided refinement becomes a practical tool in data-centric machine learning.

Where Pith is reading between the lines

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

  • Extending this to other modalities like images or text could broaden its use if suitable models are available.
  • If the pretrained model has biases, the data adjustments might embed those biases into the refined dataset.
  • Integrating this into standard preprocessing pipelines could reduce reliance on manual data cleaning steps.
  • Experiments with controlled synthetic noise would help isolate whether corrections target true noise or model-specific issues.

Load-bearing premise

That small gradient-driven adjustments to the input data correct genuine noise rather than introducing new artifacts or overfitting to the fixed model's current biases.

What would settle it

Observing that on datasets with known noise levels the method either fails to improve predictions or significantly distorts the data's statistical properties would challenge the central claim.

Figures

Figures reproduced from arXiv: 2511.10161 by Andr\'es Herrera-Poyatos, Francisco Herrera, Ignacio Aguilera-Martos, J. Javier Alonso-Ramos.

Figure 1
Figure 1. Figure 1: 2D noisy data before and after applying DenoGrad. [PITH_FULL_IMAGE:figures/full_fig_p008_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: DKL heatmap for the real-world static tabular RT IOT 2022 dataset. 5.1.2 Tabular Data: Comparative Performance As a result of the previous analyses, four denoising methods/frameworks remain viable for static tabular data: DAE, PCA, ResNet, and DenoGrad. Among these, DAE and PCA exhibit higher DKL and correlation differences in certain scenarios, particularly on the RT_IOT2022 and House Prices datasets. Den… view at source ↗
Figure 3
Figure 3. Figure 3: Signed ranked Bayes tests in train:denoised - test:denoised scenario confronting DenoGrad with the other valid state-of-the-art denoising alternatives in real-world and synthetic static tabular datasets. also help resolve the earlier question regarding ResNet’s denoising effectiveness: does a method that preserves a perfect R2 score, shows no correlation difference, and exhibits no DKL actually perform den… view at source ↗
Figure 4
Figure 4. Figure 4: R2 score obtained by the Interpretable AI models in the House Prices real-world static tabular dataset in all train-test scenarios applying ResNet as denoiser method. Based on the complete analysis, we conclude that DenoGrad is the most suitable denoising framework for static tabular data in most scenarios. It consistently preserves both the overall data distribution and the internal relationships between … view at source ↗
Figure 5
Figure 5. Figure 5: DKL heatmap for the real-world time series Daily Climate dataset. As in the static tabular setting, ResNet consistently achieves the lowest DKL and correlation differences across all datasets. However, this pattern raises the same doubts as before: whether such stability is due to actual denoising or simply a lack of any meaningful transformation. This will be explored in more detail in the following secti… view at source ↗
Figure 6
Figure 6. Figure 6: R2 score obtained by the Interpretable AI models in the WTH real-world time series dataset in all train-test scenarios applying ResNet as denoiser [PITH_FULL_IMAGE:figures/full_fig_p024_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Signed ranked Bayes tests in train:denoised - test:denoised scenario confronting DenoGrad with the other valid state-of-the-art denoising alternatives in real-world time series datasets [PITH_FULL_IMAGE:figures/full_fig_p024_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Signed ranked Bayes tests in train:denoised - test:denoised scenario confronting DenoGrad with the other valid state-of-the-art denoising alternatives in synthetic time series datasets. 24 [PITH_FULL_IMAGE:figures/full_fig_p024_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Signed ranked Bayes tests in train:denoised - test:noisy scenario confronting DenoGrad with every other state-of-the-art denoising alternative in real-world static tabular datasets. 27 [PITH_FULL_IMAGE:figures/full_fig_p027_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Signed ranked Bayes tests in train:denoised - test:noisy scenario confronting DenoGrad with every other state-of-the-art denoising alternative in synthetic static tabular datasets. 28 [PITH_FULL_IMAGE:figures/full_fig_p028_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Signed ranked Bayes tests in train:denoised - test:noisy scenario confronting DenoGrad with the other valid state-of-the-art denoising alternatives in real-world time series datasets. 29 [PITH_FULL_IMAGE:figures/full_fig_p029_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Signed ranked Bayes tests in train:denoised - test:noisy scenario confronting DenoGrad with the other valid state-of-the-art denoising alternatives in synthetic time series datasets. 30 [PITH_FULL_IMAGE:figures/full_fig_p030_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Absolute difference between the average values of the correlation matrices before and after applying each [PITH_FULL_IMAGE:figures/full_fig_p031_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: Absolute difference between the average values of the correlation matrices before and after applying each [PITH_FULL_IMAGE:figures/full_fig_p032_14.png] view at source ↗
Figure 15
Figure 15. Figure 15: Absolute difference between the average values of the correlation matrices before and after applying each [PITH_FULL_IMAGE:figures/full_fig_p033_15.png] view at source ↗
Figure 16
Figure 16. Figure 16: Absolute difference between the average values of the correlation matrices before and after applying each [PITH_FULL_IMAGE:figures/full_fig_p034_16.png] view at source ↗
Figure 17
Figure 17. Figure 17: The left column presents bar plots of the mean KL divergence for each denoising method, summarizing [PITH_FULL_IMAGE:figures/full_fig_p036_17.png] view at source ↗
Figure 18
Figure 18. Figure 18: The left column presents bar plots of the mean KL divergence for each denoising method, summarizing [PITH_FULL_IMAGE:figures/full_fig_p037_18.png] view at source ↗
Figure 19
Figure 19. Figure 19: The left column presents bar plots of the mean KL divergence for each denoising method, summarizing [PITH_FULL_IMAGE:figures/full_fig_p039_19.png] view at source ↗
Figure 20
Figure 20. Figure 20: The left column presents bar plots of the mean KL divergence for each denoising method, summarizing [PITH_FULL_IMAGE:figures/full_fig_p040_20.png] view at source ↗
Figure 21
Figure 21. Figure 21: R2 score obtained by the XAI models in the House Prices real-world static tabular dataset in all train-test scenarios applying the four denoising algorithms that passed the previous filter: DAE, DenoGrad, PCA and ResNet. 41 [PITH_FULL_IMAGE:figures/full_fig_p041_21.png] view at source ↗
Figure 22
Figure 22. Figure 22: R2 score obtained by the XAI models in the Lattice Physics real-world static tabular dataset in all train-test scenarios applying the four denoising algorithms that passed the previous filter: DAE, DenoGrad, PCA and ResNet. 42 [PITH_FULL_IMAGE:figures/full_fig_p042_22.png] view at source ↗
Figure 23
Figure 23. Figure 23: R2 score obtained by the XAI models in the Parkinsons real-world static tabular dataset in all train-test scenarios applying the four denoising algorithms that passed the previous filter: DAE, DenoGrad, PCA and ResNet. 43 [PITH_FULL_IMAGE:figures/full_fig_p043_23.png] view at source ↗
Figure 24
Figure 24. Figure 24: R2 score obtained by the XAI models in the RT IOT 2022 real-world static tabular dataset in all train-test scenarios applying the four denoising algorithms that passed the previous filter: DAE, DenoGrad, PCA and ResNet. 44 [PITH_FULL_IMAGE:figures/full_fig_p044_24.png] view at source ↗
Figure 25
Figure 25. Figure 25: R2 score obtained by the XAI models in the SUPPORT2 real-world static tabular dataset in all train-test scenarios applying the four denoising algorithms that passed the previous filter: DAE, DenoGrad, PCA and ResNet. 45 [PITH_FULL_IMAGE:figures/full_fig_p045_25.png] view at source ↗
Figure 26
Figure 26. Figure 26: R2 score obtained by the XAI models in the 2D synthetic static tabular dataset in all train-test scenarios applying the four denoising algorithms that passed the previous filter: DAE, DenoGrad, PCA and ResNet. 46 [PITH_FULL_IMAGE:figures/full_fig_p046_26.png] view at source ↗
Figure 27
Figure 27. Figure 27: R2 score obtained by the XAI models in the 3D synyhetic static tabular dataset in all train-test scenarios applying the four denoising algorithms that passed the previous filter: DAE, DenoGrad, PCA and ResNet. 47 [PITH_FULL_IMAGE:figures/full_fig_p047_27.png] view at source ↗
Figure 28
Figure 28. Figure 28: R2 score obtained by the XAI models in the Daily Climate real-world time series dataset in all train-test scenarios applying the six denoising algorithms that passed the previous filter: DAE, DenoGrad, EMD, MA, PCA and ResNet. 48 [PITH_FULL_IMAGE:figures/full_fig_p048_28.png] view at source ↗
Figure 29
Figure 29. Figure 29: R2 score obtained by the XAI models in the ECL real-world time series dataset in all train-test scenarios applying the six denoising algorithms that passed the previous filter: DAE, DenoGrad, EMD, MA, PCA and ResNet. 49 [PITH_FULL_IMAGE:figures/full_fig_p049_29.png] view at source ↗
Figure 30
Figure 30. Figure 30: R2 score obtained by the XAI models in the ETT real-world time series dataset in all train-test scenarios applying the six denoising algorithms that passed the previous filter: DAE, DenoGrad, EMD, MA, PCA and ResNet. 50 [PITH_FULL_IMAGE:figures/full_fig_p050_30.png] view at source ↗
Figure 31
Figure 31. Figure 31: R2 score obtained by the XAI models in the Microsoft Stock real-world time series dataset in all train-test scenarios applying the six denoising algorithms that passed the previous filter: DAE, DenoGrad, EMD, MA, PCA and ResNet. 51 [PITH_FULL_IMAGE:figures/full_fig_p051_31.png] view at source ↗
Figure 32
Figure 32. Figure 32: R2 score obtained by the XAI models in the WTH real-world time series dataset in all train-test scenarios applying the six denoising algorithms that passed the previous filter: DAE, DenoGrad, EMD, MA, PCA and ResNet. 52 [PITH_FULL_IMAGE:figures/full_fig_p052_32.png] view at source ↗
Figure 33
Figure 33. Figure 33: R2 score obtained by the XAI models in the Linear Combination synthetic time series dataset in all train-test scenarios applying the six denoising algorithms that passed the previous filter: DAE, DenoGrad, EMD, MA, PCA and ResNet. 53 [PITH_FULL_IMAGE:figures/full_fig_p053_33.png] view at source ↗
Figure 34
Figure 34. Figure 34: R2 score obtained by the XAI models in the Sinus synthetic time series dataset in all train-test scenarios applying the six denoising algorithms that passed the previous filter: DAE, DenoGrad, EMD, MA, PCA and ResNet. Acknowledgements This publication is part of the Project “Ethical, Responsible and General Purpose Artificial Intelligence: Applications In Risk Scenarios” (IAFER) Exp.:TSI-100927-2023-1 fun… view at source ↗
read the original abstract

In the Data-Centric Artificial Intelligence (AI) paradigm, improving data quality is essential for robust machine learning. However, many denoising methods rely on rigid statistical assumptions or require clean reference data, which limits their applicability in real-world scenarios. In this work, we propose DenoGrad, a gradient-based framework for data refinement that leverages a pretrained neural network to iteratively correct noisy observations by optimizing the input space while keeping the model fixed. DenoGrad is applicable to both tabular regression and time-series forecasting, and incorporates a consensus-based strategy to ensure temporally coherent updates in sequential settings. Experiments on ten real-world datasets show that the proposed approach yields consistent improvements in downstream predictive performance while preserving the statistical structure of the data, as measured by distributional and correlation-based metrics. In addition, DenoGrad can improve generalization in nominally clean datasets, acting as a form of dataset-level regularization. These results support model-guided data refinement as a practical component of data-centric machine learning workflows. Code is available at: https://github.com/ari-dasci/S-DenoGrad.

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

3 major / 3 minor

Summary. The paper proposes DenoGrad, a gradient-based framework for refining noisy tabular and time-series data. A pretrained neural network is held fixed while input observations are iteratively adjusted via gradients to minimize loss; a consensus mechanism enforces temporal coherence for sequential data. Experiments on ten real-world datasets report consistent gains in downstream predictive performance together with preservation of distributional and correlation-based statistical structure. The method is also claimed to improve generalization on nominally clean data by acting as dataset-level regularization. Code is released at the provided GitHub repository.

Significance. If the reported gains prove robust and independent of the fixed model's inductive biases, the approach would offer a practical, reference-free tool for data-centric workflows in regression and forecasting. The explicit attention to statistical-structure preservation and the public code release are positive features that support reproducibility and further testing. However, the significance remains provisional until the central claim—that gradient updates correct genuine noise rather than model-specific artifacts—is more rigorously isolated from confounding factors.

major comments (3)
  1. [§5] §5 (Experiments): the headline claim of 'consistent improvements across ten datasets' is presented without statistical significance tests, confidence intervals, or a clear description of baseline selection and hyperparameter protocols. This omission prevents assessment of whether the gains exceed what could arise from selection effects or favorable tuning.
  2. [§4.1] §4.1 (Method): the optimization keeps the network fixed and updates inputs to reduce loss, yet no diagnostic is supplied to distinguish recovery of true signal from alignment with the model's current decision boundaries. Because downstream evaluation uses models likely similar to the fixed one, the reported gains do not yet rule out overfitting to model-specific biases.
  3. [§4.2] §4.2 (Consensus strategy): the temporal-coherence mechanism is introduced to avoid incoherent updates, but the manuscript provides neither an ablation removing the consensus step nor quantitative metrics showing its effect on both predictive performance and correlation preservation.
minor comments (3)
  1. [§3] Notation for the gradient step size and iteration count should be introduced once in §3 and used consistently thereafter; currently the symbols appear only in the experimental protocol.
  2. [Figures] Figure captions for the distributional and correlation plots should explicitly state the exact metrics (e.g., Wasserstein distance, Pearson correlation) and the number of Monte-Carlo runs used to generate error bars.
  3. [Code availability] The GitHub repository is welcome, but the paper should list the precise hyperparameter values and random seeds employed for each dataset so that the released code can be verified without additional reverse-engineering.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive and detailed feedback. We address each major comment below and describe the revisions we will implement to improve the manuscript's rigor and clarity.

read point-by-point responses

  1. Referee: §5 (Experiments): the headline claim of 'consistent improvements across ten datasets' is presented without statistical significance tests, confidence intervals, or a clear description of baseline selection and hyperparameter protocols. This omission prevents assessment of whether the gains exceed what could arise from selection effects or favorable tuning.

    Authors: We agree that additional statistical analysis is needed to support the claims. In the revised manuscript we will add paired statistical tests (t-tests or Wilcoxon signed-rank) with p-values across the ten datasets, report 95% confidence intervals on the performance deltas, and expand the experimental protocol section to explicitly describe baseline selection criteria and hyperparameter search procedures for both the fixed refinement model and all downstream evaluators. These changes will allow readers to evaluate whether the gains are robust to selection or tuning effects. revision: yes

  2. Referee: §4.1 (Method): the optimization keeps the network fixed and updates inputs to reduce loss, yet no diagnostic is supplied to distinguish recovery of true signal from alignment with the model's current decision boundaries. Because downstream evaluation uses models likely similar to the fixed one, the reported gains do not yet rule out overfitting to model-specific biases.

    Authors: This is a substantive concern. While the original experiments already evaluate refined data on multiple downstream models, we will add a dedicated diagnostic subsection in the revision. This will include (i) cross-architecture evaluation using models whose inductive biases differ from the fixed refinement network and (ii) quantitative comparison of input-feature shifts and decision-boundary alignment before and after refinement. These additions will help separate genuine signal recovery from model-specific alignment. revision: yes

  3. Referee: §4.2 (Consensus strategy): the temporal-coherence mechanism is introduced to avoid incoherent updates, but the manuscript provides neither an ablation removing the consensus step nor quantitative metrics showing its effect on both predictive performance and correlation preservation.

    Authors: We accept that an explicit ablation is required. In the revised manuscript we will include an ablation study on the time-series datasets that compares full DenoGrad against the variant without the consensus step. We will report both predictive performance metrics (e.g., MSE, MAE) and statistical-structure metrics (Pearson correlations, distributional distances) for both variants, thereby quantifying the consensus mechanism's contribution to coherence and accuracy. revision: yes

Circularity Check

0 steps flagged

No significant circularity in the derivation chain

full rationale

The paper introduces DenoGrad as a gradient-based data refinement procedure that keeps a pretrained model fixed while iteratively adjusting inputs to reduce loss, with a consensus strategy for time-series coherence. All central claims rest on empirical results from experiments on ten real-world datasets, reporting downstream performance gains and preservation of distributional/correlation metrics. No equations or steps reduce by construction to the method's own fitted values or inputs; there are no self-definitional relations, fitted parameters renamed as predictions, or load-bearing self-citations that would force the reported improvements. The derivation and validation are self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The method rests on the premise that a pretrained model supplies reliable gradient signals for identifying and correcting noise; no explicit free parameters or invented entities are named in the abstract, though practical implementation will require choices for step size and iteration count.

free parameters (1)
  • gradient step size and iteration count
    These control how far and how many times inputs are adjusted; they must be selected to achieve refinement without excessive distortion.
axioms (1)
  • domain assumption A pretrained model on the noisy data still produces gradients that point toward useful corrections rather than amplifying existing errors.
    The framework keeps the model fixed and relies on its gradient information to guide data changes.

pith-pipeline@v0.9.0 · 5498 in / 1293 out tokens · 35329 ms · 2026-05-17T22:43:13.538704+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

79 extracted references · 79 canonical work pages · 1 internal anchor

  1. [1]

    Sidahmed, Attribute noise-sensitivity impact: Model performance and feature ranking, in: Americas Confer- ence on Information Systems, 2008, p

    M. Sidahmed, Attribute noise-sensitivity impact: Model performance and feature ranking, in: Americas Confer- ence on Information Systems, 2008, p. no info. URLhttps://api.semanticscholar.org/CorpusID:2960577

    work page 2008

  2. [2]

    Hasan, C

    R. Hasan, C. H. Chu, Drn: Detection and removal of noisy instances with self organizing map, in: International Conferences on Pattern Recognition and Artificial Intelligence, 2022, p. no info. URLhttps://api.semanticscholar.org/CorpusID:249318257

    work page 2022

  3. [3]

    Y . Li, H. Han, S. Shan, X. Chen, Disc: Learning from noisy labels via dynamic instance-specific selection and correction, 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2023) 24070– 24079. URLhttps://api.semanticscholar.org/CorpusID:260068694

    work page 2023

  4. [4]

    Papenmeier, D

    A. Papenmeier, D. Kern, G. Englebienne, C. Seifert, It’s complicated: The relationship between user trust, model accuracy and explanations in ai, ACM Trans. Comput.-Hum. Interact. (2022).doi:10.1145/3495013. URLhttps://doi.org/10.1145/3495013

    work page doi:10.1145/3495013 2022

  5. [5]

    Mokari, S

    A. Mokari, S. Eiserloh, O. Ryabchykov, U. Neugebauer, T. Bocklitz, A comparative study of robustness to noise and interpretability in u-net-based denoising of raman spectra., Spectrochimica acta. Part A, Molecular and biomolecular spectroscopy 344 Pt 1 (2025) 126577. URLhttps://api.semanticscholar.org/CorpusID:279965738

    work page 2025

  6. [6]

    Z. He, Y . Wang, Y . Yang, P. Sun, L. Wu, H. Bai, J. Gong, R. Hong, M. Zhang, Double correction framework for denoising recommendation, in: Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, 2024, p. 1062–1072.doi:10.1145/3637528.3671692. URLhttps://doi.org/10.1145/3637528.3671692

    work page doi:10.1145/3637528.3671692 2024

  7. [7]

    Kausik, K

    C. Kausik, K. Srivastava, R. Sonthalia, Double descent and overfitting under noisy inputs and distribution shift for linear denoisers, in: arxiv, 2023, p. no info. URLhttps://api.semanticscholar.org/CorpusID:258959152

    work page 2023

  8. [8]

    Larsen-Ledet, Embracing data noise, in: microsoft, 2023, p

    I. Larsen-Ledet, Embracing data noise, in: microsoft, 2023, p. no info. URLhttps://api.semanticscholar.org/CorpusID:259300333

    work page 2023

  9. [9]

    Hasan, C

    R. Hasan, C. H. Chu, Noise in datasets: What are the impacts on classification performance?, in: Proceedings of the 11th International Conference on Pattern Recognition Applications and Methods - ICPRAM, INSTICC, SciTePress, 2022, pp. 163–170.doi:10.5220/0010782200003122

    work page doi:10.5220/0010782200003122 2022

  10. [10]

    García-Gil, J

    D. García-Gil, J. Luengo, S. García, F. Herrera, Enabling smart data: noise filtering in big data classification, Information Sciences 479 (2019) 135–152

    work page 2019

  11. [11]

    K. W. Al-Sabbagh, M. Staron, R. Hebig, Improving test case selection by handling class and attribute noise, J. Syst. Softw. 183 (2021) 111093. URLhttps://api.semanticscholar.org/CorpusID:244245120

    work page 2021

  12. [12]

    R. K. L. Kennedy, J. M. Johnson, T. M. Khoshgoftaar, The effects of class label noise on highly-imbalanced big data, 2021 IEEE 33rd International Conference on Tools with Artificial Intelligence (ICTAI) (2021) 1427–1433. URLhttps://api.semanticscholar.org/CorpusID:245388536

    work page 2021

  13. [13]

    E, Statistical physics, plasmas, fluids, and related interdisciplinary topics 48 1 (1993) R13–R16

    Schreiber, Determination of the noise level of chaotic time series., Physical review. E, Statistical physics, plasmas, fluids, and related interdisciplinary topics 48 1 (1993) R13–R16. URLhttps://api.semanticscholar.org/CorpusID:12164904

    work page 1993

  14. [14]

    J. M. Johnson, T. M. Khoshgoftaar, A survey on classifying big data with label noise, ACM Journal of Data and Information Quality 14 (2022) 1 – 43. URLhttps://api.semanticscholar.org/CorpusID:248070196

    work page 2022

  15. [15]

    Pontil, S

    M. Pontil, S. Mukherjee, F. Girosi, On the noise model of support vector machines regression, in: International Conference on Algorithmic Learning Theory, 2000, p. no info. URLhttps://api.semanticscholar.org/CorpusID:6244043

    work page 2000

  16. [16]

    C. M. Bishop, C. S. Quazaz, Regression with input-dependent noise: A bayesian treatment, in: Neural Information Processing Systems, 1996, p. no info. URLhttps://api.semanticscholar.org/CorpusID:14474914

    work page 1996

  17. [17]

    P. W. Goldberg, C. K. I. Williams, C. M. Bishop, Regression with input-dependent noise: A gaussian process treatment, in: Neural Information Processing Systems, 1997, p. no info. URLhttps://api.semanticscholar.org/CorpusID:7482528 55 arXivTemplateA PREPRINT

    work page 1997

  18. [18]

    Shankar, D

    S. Shankar, D. Sheldon, T. Sun, J. Pickering, T. G. Dietterich, Three-quarter sibling regression for denoising observational data, in: Proceedings of the Twenty-Eighth International Joint Conference on Artificial Intelligence, IJCAI-19, International Joint Conferences on Artificial Intelligence Organization, 2019, pp. 5960–5966. doi: 10.24963/ijcai.2019/8...

    work page doi:10.24963/ijcai.2019/826 2019

  19. [19]

    Chaudhary, I

    P. Chaudhary, I. Chauhan, B. R. Manoj, M. R. Bhatnagar, Linear regression-based channel estimation for non- gaussian noise, 2024 IEEE 99th Vehicular Technology Conference (VTC2024-Spring) (2024) 1–6. URLhttps://api.semanticscholar.org/CorpusID:272858899

    work page 2024

  20. [20]

    S. G. Kwak, J. H. Kim, Central limit theorem: the cornerstone of modern statistics, Korean journal of anesthesiol- ogy 70 (2) (2017) 144

    work page 2017

  21. [21]

    Santhosh, P

    S. Santhosh, P. Soori, V . Shashidhar, Impact of additive noise on information loss characteristics of datasets, 2023 International Conference on the Confluence of Advancements in Robotics, Vision and Interdisciplinary Technology Management (IC-RVITM) (2023) 1–8. URLhttps://api.semanticscholar.org/CorpusID:267772787

    work page 2023

  22. [22]

    Barnett, O

    E. Barnett, O. Kaiser, J. Masci, E. C. Wit, S. Fulda, Generative models for periodicity detection in noisy signals, Clocks & Sleep 6 (3) (2024) 359–388

    work page 2024

  23. [23]

    Jafarigol, T

    E. Jafarigol, T. Trafalis, The paradox of noise: An empirical study of noise-infusion mechanisms to improve generalization, stability, and privacy in federated learning (2023).arXiv:2311.05790. URLhttps://arxiv.org/abs/2311.05790

    work page arXiv 2023

  24. [24]

    Gupta, A

    S. Gupta, A. Gupta, Dealing with noise problem in machine learning data-sets: A systematic review, Procedia Computer Science 161 (2019) 466–474

    work page 2019

  25. [25]

    S. Wang, H. Chen, A novel deep learning method for the classification of power quality disturbances using deep convolutional neural network, Applied energy 235 (2019) 1126–1140

    work page 2019

  26. [26]

    Explainable artificial intelligence (xai): Concepts, taxonomies, opportunities and challenges toward responsi- ble ai,

    A. Barredo Arrieta, N. Díaz-Rodríguez, J. Del Ser, A. Bennetot, S. Tabik, A. Barbado, S. Garcia, S. Gil- Lopez, D. Molina, R. Benjamins, R. Chatila, F. Herrera, Explainable artificial intelligence (XAI): Concepts, taxonomies, opportunities and challenges toward responsible AI, Information Fusion 58 (2020) 82–115. doi: 10.1016/j.inffus.2019.12.012

    work page doi:10.1016/j.inffus.2019.12.012 2020

  27. [27]

    Gunning, Explainable artificial intelligence (xai), Defense advanced research projects agency (DARPA), nd Web 2 (2) (2017) 1

    D. Gunning, Explainable artificial intelligence (xai), Defense advanced research projects agency (DARPA), nd Web 2 (2) (2017) 1

    work page 2017

  28. [28]

    Towards A Rigorous Science of Interpretable Machine Learning

    F. Doshi-Velez, B. Kim, Towards a rigorous science of interpretable machine learning, arXiv preprint arXiv:1702.08608 (2017)

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  29. [29]

    A. E. Hoerl, R. W. Kennard, Ridge regression: Biased estimation for nonorthogonal problems, Technometrics 12 (1) (1970) 55–67

    work page 1970

  30. [30]

    S. Wold, A. Ruhe, H. Wold, W. Dunn, Iii, The collinearity problem in linear regression. the partial least squares (pls) approach to generalized inverses, SIAM Journal on Scientific and Statistical Computing 5 (3) (1984) 735–743

    work page 1984

  31. [31]

    Breiman, J

    L. Breiman, J. Friedman, R. A. Olshen, C. J. Stone, Classification and regression trees, Routledge, 2017

    work page 2017

  32. [32]

    Drucker, C

    H. Drucker, C. J. Burges, L. Kaufman, A. Smola, V . Vapnik, Support vector regression machines, Advances in neural information processing systems 9 (1996)

    work page 1996

  33. [33]

    E. Fix, J. L. Hodges Jr, Discriminatory analysis. nonparametric discrimination: Consistency properties, Tech. rep., USA Air Force School of Aviation Medicine, Randolph Field (1951)

    work page 1951

  34. [34]

    G. E. Box, G. M. Jenkins, Time series analysis: Forecasting and control, Holden-Day, 1970

    work page 1970

  35. [35]

    Crook, M

    B. Crook, M. Schlüter, T. Speith, Revisiting the performance-explainability trade-off in explainable artificial intelligence (xai), in: 2023 IEEE 31st International Requirements Engineering Conference Workshops (REW), IEEE, 2023, pp. 316–324

    work page 2023

  36. [36]

    Peracchio, G

    L. Peracchio, G. Nicora, T. M. Buonocore, R. Bellazzi, E. Parimbelli, Do you trust your model explanations? an analysis of xai performance under dataset shift, in: International Conference on Artificial Intelligence in Medicine, Springer, 2024, pp. 257–266

    work page 2024

  37. [37]

    Milanfar, M

    P. Milanfar, M. Delbracio, Denoising: a powerful building block for imaging, inverse problems and machine learning, Philosophical transactions. Series A, Mathematical, physical, and engineering sciences 383 (2024). URLhttps://api.semanticscholar.org/CorpusID:272550819

    work page 2024

  38. [38]

    Advancing humanoid locomotion: Mastering challenging terrains with denoising world model learning.arXiv preprint arXiv:2408.14472, 2024

    X. Gu, Y .-J. Wang, X. Zhu, C. Shi, Y . Guo, Y . Liu, J. Chen, Advancing humanoid locomotion: Mastering challenging terrains with denoising world model learning (2024).arXiv:2408.14472. URLhttps://arxiv.org/abs/2408.14472 56 arXivTemplateA PREPRINT

    work page arXiv 2024

  39. [39]

    Frusque, O

    G. Frusque, O. Fink, Robust time series denoising with learnable wavelet packet transform, Advanced Engineering Informatics 62 (2024) 102669

    work page 2024

  40. [40]

    Makwana, G

    D. Makwana, G. Deshmukh, O. Susladkar, S. Mittal, et al., Livenet: A novel network for real-world low-light image denoising and enhancement, in: Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, 2024, pp. 5856–5865

    work page 2024

  41. [41]

    P. Zhao, W. Zhang, X. Cao, X. Li, Denoising diffusion probabilistic model-enabled data augmentation method for intelligent machine fault diagnosis, Engineering Applications of Artificial Intelligence 139 (2025) 109520

    work page 2025

  42. [42]

    L. P. F. Garcia, A. C. Lorena, A. C. P. de Leon Ferreira de Carvalho, A study on class noise detection and elimination, 2012 Brazilian Symposium on Neural Networks (2012) 13–18. URLhttps://api.semanticscholar.org/CorpusID:14146586

    work page 2012

  43. [43]

    S. Zhou, Z. He, X. Chen, W. Chang, An anomaly detection method for uav based on wavelet decomposition and stacked denoising autoencoder, Aerospace 11 (5) (2024) 393

    work page 2024

  44. [44]

    García, J

    S. García, J. Luengo, F. Herrera, Dealing with noisy data, in: Dealing with Noisy Data, 2015, p. no info. URLhttps://api.semanticscholar.org/CorpusID:268092955

    work page 2015

  45. [45]

    J. Shi, K. Zhang, C. Guo, Y . Yang, Y . Xu, J. Wu, A survey of label-noise deep learning for medical image analysis, Medical Image Analysis 95 (2024) 103166.doi:https://doi.org/10.1016/j.media.2024.103166

    work page doi:10.1016/j.media.2024.103166 2024

  46. [46]

    Majumdar, Blind denoising autoencoder, IEEE transactions on neural networks and learning systems 30 (1) (2018) 312–317

    A. Majumdar, Blind denoising autoencoder, IEEE transactions on neural networks and learning systems 30 (1) (2018) 312–317

    work page 2018

  47. [47]

    Boudraa, J.-C

    A.-O. Boudraa, J.-C. Cexus, et al., Denoising via empirical mode decomposition, Proc. IEEE ISCCSP 4 (2006) (2006)

    work page 2006

  48. [48]

    de Vries, A

    I. de Vries, A. de Jong, M. van der Hout-van der Jagt, J. van Laar, R. Vullings, Deep learning for sparse domain kalman filtering with applications on ecg denoising and motility estimation, IEEE Transactions on Biomedical Engineering 71 (8) (2024) 2321–2329.doi:10.1109/TBME.2024.3368105

    work page doi:10.1109/tbme.2024.3368105 2024

  49. [49]

    Z. Shan, J. Yang, M. A. Sanjuán, C. Wu, H. Liu, A novel adaptive moving average method for signal denoising in strong noise background, The European Physical Journal Plus 137 (1) (2022) 50

    work page 2022

  50. [50]

    W. Dong, M. Wo´ zniak, J. Wu, W. Li, Z. Bai, Denoising aggregation of graph neural networks by using principal component analysis, IEEE Transactions on Industrial Informatics 19 (3) (2023) 2385–2394. doi:10.1109/TII. 2022.3156658

    work page doi:10.1109/tii 2023

  51. [51]

    H. Kang, C. Park, H. Yang, Evaluation of denoising performance of resnet deep learning model for ultrasound images corresponding to two frequency parameters, Bioengineering 11 (7) (2024) 723

    work page 2024

  52. [52]

    C. Tian, M. Zheng, W. Zuo, B. Zhang, Y . Zhang, D. Zhang, Multi-stage image denoising with the wavelet transform, Pattern Recognition 134 (2023) 109050

    work page 2023

  53. [53]

    D. L. Donoho, De-noising by soft-thresholding, IEEE transactions on information theory 41 (3) (2002) 613–627

    work page 2002

  54. [54]

    W. S. McCulloch, W. Pitts, A logical calculus of the ideas immanent in nervous activity, The bulletin of mathemat- ical biophysics 5 (1943) 115–133

    work page 1943

  55. [55]

    Z. L. Liu, Artificial neural networks, in: Artificial Intelligence for Engineers: Basics and Implementations, Springer, 2025, pp. 175–190

    work page 2025

  56. [56]

    Robbins, S

    H. Robbins, S. Monro, A stochastic approximation method, The annals of mathematical statistics (1951) 400–407

    work page 1951

  57. [57]

    K. Cho, B. van Merriënboer, C. Gulcehre, D. Bahdanau, F. Bougares, H. Schwenk, Y . Bengio, Learning phrase representations using RNN encoder–decoder for statistical machine translation, in: A. Moschitti, B. Pang, W. Daelemans (Eds.), Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP), Association for Computation...

    work page doi:10.3115/v1/ 2014

  58. [58]

    Pascanu, T

    R. Pascanu, T. Mikolov, Y . Bengio, On the difficulty of training recurrent neural networks, in: International conference on machine learning, Pmlr, 2013, pp. 1310–1318

    work page 2013

  59. [59]

    S. Hochreiter, The vanishing gradient problem during learning recurrent neural nets and problem solutions, International Journal of Uncertainty, Fuzziness and Knowledge-Based Systems 6 (02) (1998) 107–116

    work page 1998

  60. [60]

    V . Nair, G. E. Hinton, Rectified linear units improve restricted boltzmann machines, in: Proceedings of the 27th international conference on machine learning (ICML-10), 2010, pp. 807–814

    work page 2010

  61. [61]

    K. He, X. Zhang, S. Ren, J. Sun, Deep residual learning for image recognition, in: Proceedings of the IEEE conference on computer vision and pattern recognition, 2016, pp. 770–778. 57 arXivTemplateA PREPRINT

    work page 2016

  62. [62]

    Camuto, M

    A. Camuto, M. Willetts, U. Simsekli, S. J. Roberts, C. C. Holmes, Explicit regularisation in gaussian noise injections, Advances in Neural Information Processing Systems 33 (2020) 16603–16614

    work page 2020

  63. [63]

    Hochreiter, J

    S. Hochreiter, J. Schmidhuber, Long short-term memory, Neural computation 9 (8) (1997) 1735–1780

    work page 1997

  64. [64]

    M. Beck, K. Pöppel, M. Spanring, A. Auer, O. Prudnikova, M. Kopp, G. Klambauer, J. Brandstetter, S. Hochreiter, xlstm: Extended long short-term memory, Advances in Neural Information Processing Systems 37 (2024) 107547–107603

    work page 2024

  65. [65]

    J. J. Hopfield, Neural networks and physical systems with emergent collective computational abilities., Proceedings of the national academy of sciences 79 (8) (1982) 2554–2558

    work page 1982

  66. [66]

    Svozil, V

    D. Svozil, V . Kvasnicka, J. Pospichal, Introduction to multi-layer feed-forward neural networks, Chemometrics and intelligent laboratory systems 39 (1) (1997) 43–62

    work page 1997

  67. [67]

    van Doorn, A

    J. van Doorn, A. Ly, M. Marsman, E.-J. Wagenmakers, Bayesian rank-based hypothesis testing for the rank sum test, the signed rank test, and spearman’sρ, Journal of Applied Statistics 47 (16) (2020) 2984–3006

    work page 2020

  68. [68]

    harlfoxem, House Sales in King County, USA, Kaggle, https://www.kaggle.com/datasets/harlfoxem/housesalesprediction (2014)

    work page 2014

  69. [69]

    Huu Tiep, Lattice-physics (PWR fuel assembly neutronics simulation results), UCI Machine Learning Reposi- tory, DOI: https://doi.org/10.24432/C5BK64 (2024)

    N. Huu Tiep, Lattice-physics (PWR fuel assembly neutronics simulation results), UCI Machine Learning Reposi- tory, DOI: https://doi.org/10.24432/C5BK64 (2024)

    work page doi:10.24432/c5bk64 2024

  70. [70]

    Tsanas, M

    A. Tsanas, M. Little, Parkinsons Telemonitoring, UCI Machine Learning Repository, DOI: https://doi.org/10.24432/C5ZS3N (2009)

    work page doi:10.24432/c5zs3n 2009

  71. [71]

    B. S., R. Nagapadma, RT-IoT2022 , UCI Machine Learning Repository, DOI: https://doi.org/10.24432/C5P338 (2023)

    work page doi:10.24432/c5p338 2023

  72. [72]

    Harrel, SUPPORT2, UCI Machine Learning Repository, DOI: https://doi.org/10.3886/ICPSR02957.v2 (2023)

    F. Harrel, SUPPORT2, UCI Machine Learning Repository, DOI: https://doi.org/10.3886/ICPSR02957.v2 (2023)

    work page doi:10.3886/icpsr02957.v2 2023

  73. [73]

    W. U. API, Daily Climate time series data, Kaggle, URL: https://www.kaggle.com/datasets/sumanthvrao/daily- climate-time-series-data (2019)

    work page 2019

  74. [74]

    H. Zhou, S. Zhang, J. Peng, S. Zhang, J. Li, H. Xiong, W. Zhang, Informer: Beyond efficient transformer for long sequence time-series forecasting, in: The Thirty-Fifth AAAI Conference on Artificial Intelligence, AAAI 2021, Virtual Conference, V ol. 35, AAAI Press, 2021, pp. 11106–11115

    work page 2021

  75. [75]

    V . V . Venkitesh, Microsoft Stock- Time Series Analysis, Kaggle, URL: https://www.kaggle.com/datasets/vijayvvenkitesh/microsoft-stock-time-series-analysis (2021)

    work page 2021

  76. [76]

    Kopsinis, S

    Y . Kopsinis, S. McLaughlin, Development of emd-based denoising methods inspired by wavelet thresholding, IEEE Transactions on signal Processing 57 (4) (2009) 1351–1362

    work page 2009

  77. [77]

    Terrien, C

    J. Terrien, C. Marque, B. Karlsson, Automatic detection of mode mixing in empirical mode decomposition using non-stationarity detection: application to selecting imfs of interest and denoising, EURASIP Journal on Advances in Signal Processing 2011 (2011) 1–8

    work page 2011

  78. [78]

    Boudraa, J.-C

    A.-O. Boudraa, J.-C. Cexus, Z. Saidi, Emd-based signal noise reduction, International Journal of Signal Processing 1 (1) (2004) 33–37

    work page 2004

  79. [79]

    Daubechies, The wavelet transform, time-frequency localization and signal analysis, IEEE transactions on information theory 36 (5) (2002) 961–1005

    I. Daubechies, The wavelet transform, time-frequency localization and signal analysis, IEEE transactions on information theory 36 (5) (2002) 961–1005. 58

    work page 2002