Pith. sign in

REVIEW 4 major objections 5 minor 22 references

Wavelet-Enhanced PaDiM for Industrial Anomaly Detection

T0 review · 4 major / 5 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read Replacing PaDiM's random channel selection with wavelet frequency subbands keeps detection accuracy and adds interpretability.

desk verdict A plausible, clearly described wavelet-channel-selection plug-in for PaDiM, but the headline AUCs are per-class test-set-optimized oracles, not the performance of a single system. read the letter →

arxiv 2508.16034 v1 pith:UFO7LUFH submitted 2025-08-22 cs.CV

Cory Gardner , Byungseok Min , Tae-Hyuk Ahn This is my paper
classification cs.CV
keywords anomalydetectionlocalizationwavelettransformPaDiMfeatureselectionMVTecADindustrialinspection
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

WE-PaDiM claims that a structured, frequency-based feature selection can replace the random channel sampling at the heart of PaDiM without losing accuracy. The authors apply a two-dimensional discrete wavelet transform to each layer's CNN feature map, keep chosen subbands (low-frequency LL, horizontal LH, vertical HL), align them, concatenate them, and then fit PaDiM's per-patch multivariate Gaussians. On the MVTec AD benchmark, with per-class tuned configurations, the method reports an average image-level AUC of 99.32% and pixel-level AUC of 92.10%, competitive with established embedding-based detectors. The point of the work is not just accuracy: choosing subbands by frequency content gives practitioners an interpretable handle on what the model attends to (coarse structure vs. fine edges), which random selection cannot provide.

What carries the argument

The central object is the DWT-before-concatenation pipeline: per-layer 2D DWT, subband selection S, spatial alignment, channel-wise concatenation, then PaDiM's per-patch multivariate Gaussian modeling. The load-bearing idea is that frequency content is a meaningful organizing principle: LL carries coarse/global information, LH/HL/HH carry directional detail, so selecting subbands is a structured alternative to random channel sampling. The Mahalanobis distance to the fitted normal distribution produces both the pixel-level anomaly map and the image-level score.

What would settle it

Run WE-PaDiM with a fixed configuration (e.g., Haar, level 1, subbands LL) on all 15 MVTec classes and compare the average Image and Pixel AUC to the per-class optimized numbers. If the fixed-configuration average is substantially lower, the headline results are mostly test-set fitting rather than robust gains.

Watch

Extended reading notes

Core claim

On its own terms, the paper establishes that DWT-before-concatenation is a viable and principled substitute for random dimensionality reduction in PaDiM. Instead of drawing a fixed-size random subset of concatenated CNN channels, WE-PaDiM decomposes each layer's feature maps into LL, LH, HL, and HH subbands, selects a subset, aligns the maps to a common resolution, and concatenates them before fitting a multivariate Gaussian per spatial location and scoring with Mahalanobis distance. Across ResNet-18 and EfficientNet B0-B6 backbones on MVTec AD, the method reaches an average of 99.32% image AUC when the best configuration is selected per class, and 92.10% pixel AUC under a separate per-class

Load-bearing premise

The headline numbers assume that picking the best configuration per class on the MVTec test set is a fair way to judge the method; if a single fixed configuration must be chosen across all classes, the reported averages likely drop.

Editorial extensions

If this is right

  • Low-frequency (LL) subbands alone can serve as a reduced, interpretable feature set for image-level detection, since they consistently lead in image AUC.
  • For localization, adding directional detail subbands (LH, HL, and usually HH) improves pixel-level accuracy, so a practitioner can tune subband selection to the expected defect type.
  • The method stays within PaDiM's training-free, Gaussian-modeling regime, so it inherits PaDiM's practical speed and memory profile while adding only DWT overhead.
  • Different backbones and categories favor different wavelets and subbands, meaning configuration choice is a meaningful axis of the method rather than a nuisance parameter.

Reading between the lines

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

  • Editorial: the 99.32% Image AUC and 92.10% Pixel AUC come from two different per-class selection passes (Table 4 vs Table 5), so they do not describe a single model; a fixed configuration would likely score lower on at least one of the metrics.
  • Editorial: because configurations were selected per class on the test set from tens to hundreds of candidates per class, the averages may overstate deployable performance; a fair comparison would hold the configuration fixed across classes or use a held-out validation split.
  • Editorial: the same DWT-before-concatenation idea could be dropped into other patch-embedding anomaly detectors, not just PaDiM, with testable consequences for localization on fine defects.
  • Editorial: an explicit head-to-head against random channel selection at matched feature dimension would isolate whether gains come from the frequency structure itself or simply from using more or different channels.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper proposes Wavelet-Enhanced PaDiM (WE-PaDiM), which replaces the random channel selection in PaDiM with a structured, frequency-domain selection based on a 2D Discrete Wavelet Transform applied independently to multi-layer CNN feature maps. Selected subbands (e.g., LL, LH, HL) are concatenated channel-wise and fed into PaDiM's multivariate Gaussian patch modeling. The method is evaluated on MVTec AD with ResNet-18 and EfficientNet B0-B6 backbones. The authors report average Image AUC of 99.32% and Pixel AUC of 92.10% under per-class optimized configurations, and provide a detailed analysis of the impact of wavelet type, subband choice, decomposition level, and other hyperparameters on detection and localization performance.

Significance. If the headline results corresponded to a fixed, deployable configuration, WE-PaDiM would be a simple, efficient, and interpretable alternative to PaDiM's random channel selection. The paper's strengths include a clear method description, a public code repository, and a systematic ablation of subband choices with significance tests and visualizations. However, the central performance claim is currently based on per-class test-set-optimized configurations, which are oracle maxima over sometimes hundreds of candidate settings. Consequently, the contribution as stated is not yet established; the paper needs a defensible evaluation protocol or a major reframing of the claims.

major comments (4)
  1. [Abstract; §4.3; §4.8; Tables 4/5] The headline numbers 99.32% Image-AUC and 92.10% Pixel-AUC are per-class test-set-optimized maxima, not the performance of a fixed WE-PaDiM system. For each MVTec class, the best configuration is selected directly on the test set from grids with as many as 430 candidates (Table 4, bottle) or 236 (carpet). This is selection on the evaluation set. The paper's own fixed-across-classes backbone configurations are substantially lower (Table 1: efficientnet-b6, Haar, J=1, LL, σ=2, ε=0.1 => 0.9884 Image / 0.8609 Pixel; Table 2: resnet18, Haar, J=1, HH_LH_LL, σ=2, ε=0.001 => 0.9077 Pixel / 0.9430 Image), and even these are selected at the backbone level on the test set. Moreover, the two headline figures come from different selection criteria: Table 4's 0.9932 Image average corresponds to only 0.8383 average Pixel AUC, while Table 5's 0.9210 Pixel average corresponds to 0.9555 average Image AUC;
  2. [§5, 'Comparison with Original PaDiM'] The paper states that 'a direct, controlled comparison with the original PaDiM's random selection ... was not the primary focus of the presented results.' This is load-bearing because the core motivation is to replace random selection with DWT-based selection. Without a controlled comparison using the same backbones, layers, and evaluation protocol (e.g., matching the number of selected dimensions DW to PaDiM's DR), the claim that WE-PaDiM is a 'competitive and interpretable alternative' to random selection is not supported. The component analysis in Fig. 3 shows within-method effects, but not how the structured selection compares with PaDiM's random channel selection on the same backbone. A fixed-configuration PaDiM baseline must be included.
  3. [§4.8, Tables 4/5] The per-class 'best' configurations are obtained under highly uneven search effort: the 'Opt. # Cfgs' column ranges from 1 (e.g., Table 5, screw) to 430 (Table 4, bottle). This means that per-class maxima are not directly comparable across classes, and classes with more searched configurations have a higher chance of a spuriously favorable maximum. Reporting per-class maxima without search-effort or variance information overstates the consistency of the method. The authors should either provide fixed-configuration results across all classes or present the per-class search as a hyperparameter optimization analysis, with the number of trials and the distribution of scores clearly reported.
  4. [§4.5, Fig. 3; Table 3] The statistical significance analysis in §4.5 is presented as evidence for the benefit of LL over detail subbands, but the comparisons are performed over all configurations in the search grid, which are not independent (they share backbones, layers, and hyperparameter families) and are selected on the test set. The p-values are therefore optimistic and should be interpreted as descriptive rather than inferential. Table 3's averages over all backbones, classes, and configurations mix very different regimes; the qualitative conclusions about LL for Image AUC and detail subbands for Pixel AUC are plausible, but the confidence intervals and test-set selection need to be acknowledged.
minor comments (5)
  1. [§3.4, Eq. (7)] Typographical error: 'regularization term ϵ. )' should read 'regularization term ϵI'.
  2. [§4.4 and §4.7] The subsection 'Impact of Subband Selection' appears twice, nearly verbatim (in §4.4 and again in §4.7). One occurrence should be removed or replaced with a cross-reference.
  3. [§2.1, reference [16]] Reference [16] is attributed to Ruff et al. as 'DifferNet's normalizing-flow modeling'; DifferNet is by Rudolph et al. (2021). Please correct the citation and the associated text.
  4. [Tables 4/5 captions] The captions state that parameters are 'unique values across tied top configurations,' but it is unclear how the single reported Image/Pixel AUC values were computed when multiple tied configurations have different secondary-metric values. State the aggregation rule (e.g., average, best secondary metric).
  5. [§5, 'Interpretability and Practical Implications'] The sentence 'The comparable computational efficiency (as suggested by the abstract, though detailed timings are pending)' is vague. Either provide concrete runtime/memory comparisons or omit the efficiency claim.

Circularity Check

1 steps flagged · score 4.0 of 10

Algorithmic derivation is self-contained and contains no self-citation chain; the partial circularity is confined to the headline performance, which is a per-class test-set-optimized oracle rather than a fixed-system result.

  1. fitted input called prediction [Abstract; Section 4.3 'Overall Performance'; Tables 4 and 5 captions/rows]
    "The proposed method achieves high performance in both anomaly detection and localization, yielding average results of approximately 99.32% Image-AUC and 92.10% Pixel-AUC across the 15 MVTec AD categories when using per-class optimized configurations."

    The headline numbers are not the output of any fixed WE-PaDiM instantiation. They are per-class maxima over a test-set grid search: Table 4 lists up to 430 candidate configurations for a single class, selected as 'the best consolidated configuration for each class' using test-set performance. A configuration is thus defined as the one that maximizes the reported test-set AUC, and reporting that AUC as the method's result is, by construction, reporting the maximizing value of the search rather than an independent estimate. The two headline metrics are also optimized under different criteria (Table 4's best-Image configs average Pixel-AUC 0.8383; Table 5's best-Pixel configs average Image-AUC 0.9555), so no single system attains both. This is fitted-input-called-prediction: the fit is the pe

full rationale

The derivation chain of the method itself is not circular. WE-PaDiM's feature construction is explicitly defined in Eqs. 2-3 (per-layer DWT, subband selection, alignment, concatenation), and the anomaly scores are the standard PaDiM multivariate-Gaussian/Mahalanobis computation in Eqs. 5-8. None of these equations assumes the reported AUCs, and no load-bearing result is imported from the authors' own prior work. The references to PaDiM [2], MVTec AD [1], and wavelets [3] are external, machine-comparable baselines/tools, not self-citations. The only circular step I can exhibit is in the evaluation protocol: the abstract's headline 99.32% Image-AUC / 92.10% Pixel-AUC is computed by selecting the best configuration per MVTec class directly on the test set, with up to hundreds of candidates per class (Tables 4/5), and the two numbers come from different optimization criteria. That makes the headline a test-set-optimized oracle rather than a fixed-model result. Section 5 also concedes that a direct controlled comparison against original PaDiM's random selection was not the primary focus, so the paper's comparative claim is under-evidenced, though this is a missing-evidence issue rather than a definitional circularity. Because the central algorithm has independent, non-circular content and the circularity is limited to the reported headline performance, a score of 4 (partial evaluation circularity) is appropriate; this is not a case where the derivation is equivalent to its inputs.

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

The method introduces no new entities: wavelet subbands are standard transforms, and the model parameters are per-patch Gaussian statistics estimated from training data. The main free choices are the per-class test-set-selected configurations and the standard PaDiM hyperparameters (sigma, epsilon, wavelet family/level).

free parameters (4)
  • per-class configuration (wavelet type, level J, subband set, sigma, cov_reg) = e.g., bottle: Eff-b5, haar/sym4, level 1/2, subbands HH_LL/LL, sigma 2/4, cov_reg 0.001-0.1; averaged best per-class con
    The stated performance is the average of the best configuration per MVTec class selected on the test set over a grid of tens to hundreds of candidates per class ('Opt. # Cfgs' in Tables 4/5). This selection is a free parameter that directly produces the headline numbers.
  • Gaussian smoothing sigma (σ) = 2.0, 4.0, 6.0
    Post-processing hyperparameter tuned per configuration; affects Pixel AUC and final anomaly map smoothness.
  • Covariance regularization epsilon (ϵ) = 0.1, 0.01, 0.001
    Added to the diagonal of per-patch covariance matrices (Eq. 7) to ensure invertibility; chosen by grid search per class/backbone.
  • Wavelet basis and decomposition level (Haar, db2, db4, sym4; J=1 or 2) = varies per class
    Defines the frequency decomposition; per-class optimal choices differ (Tables 4/5).
assumptions (4)
  • standard math The 2D DWT with an orthogonal wavelet (Haar, Daubechies) is an invertible, information-preserving decomposition whose subband coefficients faithfully represent the input feature map.
    Invoked in Section 3.2 when computing F^s_l; the method relies on DWT coefficients being a valid re-representation of Fl.
  • domain assumption Features from ImageNet pre-trained CNNs transfer to industrial anomaly detection and localization.
    Adopted from PaDiM and the embedding-based literature (Sections 2.1 and 3.1); the entire feature pipeline depends on this transfer.
  • domain assumption Anomalies in industrial images manifest with distinct frequency characteristics in CNN feature maps (fine scratches in high-frequency bands, discoloration or structural changes in low-frequency bands).
    The paper's motivating hypothesis (Section 1, 'We hypothesize...') and the basis for subband selection; not independently validated outside the MVTec AD test set.
  • domain assumption MVTec AD is a representative benchmark, and per-class test-set-optimized configurations are an acceptable way to report method performance.
    The evaluation protocol in Section 4.3 selects best configs per class on the test set; the claim of 99.32/92.10 depends on this convention being accepted.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Wavelet-Enhanced PaDiM for Industrial Anomaly Detection." pith.science (2026). https://pith.science/paper/UFO7LUFH

@misc{pith2026250816034,
  author       = {Pith},
  title        = {Pith review of: Wavelet-Enhanced PaDiM for Industrial Anomaly Detection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/UFO7LUFH}},
  note         = {Machine review of arXiv:2508.16034}
}
read the original abstract

Anomaly detection and localization in industrial images are essential for automated quality inspection. PaDiM, a prominent method, models the distribution of normal image features extracted by pre-trained Convolutional Neural Networks (CNNs) but reduces dimensionality through random channel selection, potentially discarding structured information. We propose Wavelet-Enhanced PaDiM (WE-PaDiM), which integrates Discrete Wavelet Transform (DWT) analysis with multi-layer CNN features in a structured manner. WE-PaDiM applies 2D DWT to feature maps from multiple backbone layers, selects specific frequency subbands (e.g., LL, LH, HL), spatially aligns them, and concatenates them channel-wise before modeling with PaDiM's multivariate Gaussian framework. This DWT-before-concatenation strategy provides a principled method for feature selection based on frequency content relevant to anomalies, leveraging multi-scale wavelet information as an alternative to random selection. We evaluate WE-PaDiM on the challenging MVTec AD dataset with multiple backbones (ResNet-18 and EfficientNet B0-B6). The method achieves strong performance in anomaly detection and localization, yielding average results of 99.32% Image-AUC and 92.10% Pixel-AUC across 15 categories with per-class optimized configurations. Our analysis shows that wavelet choices affect performance trade-offs: simpler wavelets (e.g., Haar) with detail subbands (HL or LH/HL/HH) often enhance localization, while approximation bands (LL) improve image-level detection. WE-PaDiM thus offers a competitive and interpretable alternative to random feature selection in PaDiM, achieving robust results suitable for industrial inspection with comparable efficiency.

Figures

Figures reproduced from arXiv: 2508.16034 by the authors.

Figure 1
Figure 1. Visualization of anomaly detection results using our WE-PaDiM method on the MVTec AD dataset. From [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Overview of the Wavelet-Enhanced PaDiM (WE-PaDiM) architecture. The diagram illustrates the WE [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
Figure 3
Figure 3. Impact of wavelet components on anomaly de [PITH_FULL_IMAGE:figures/full_fig_p010_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Class-specific performance analysis. Top row: Performance comparison between texture and object [PITH_FULL_IMAGE:figures/full_fig_p011_4.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

22 extracted references · 15 canonical work pages

  1. [1]

    Mvtec ad–a comprehensive real- world dataset for unsupervised anomaly detection

    Paul Bergmann, Michael Fauser, David Sattlegger, and Carsten Steger. Mvtec ad–a comprehensive real- world dataset for unsupervised anomaly detection. Proceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition, pages 9592– 9600, 2019. doi: 10.1109/CVPR.2019.00982

    arXiv 2019

  2. [2]

    Padim: A patch distribu- tion modeling framework for anomaly detection and localization

    Thomas Defard, Aleksandr Setkov, Angelique Loesch, and Romaric Audigier. Padim: A patch distribu- tion modeling framework for anomaly detection and localization. In Proceedings of the International Con- ference on Pattern Recognition Workshops (ICPR Workshops), pages 475–489, 2020. doi: 10.48550/ arXiv.2011.08785

    arXiv doi:10.48550/arxiv.2011.08785 2020

  3. [3]

    A theory for multiresolution signal decomposition: The wavelet representation

    St’ephane Mallat. A theory for multiresolution signal decomposition: The wavelet representation. IEEE Transactions on Pattern Analysis and Machine Intel- ligence, 11(7):674–693, 1989

    work page 1989

  4. [4]

    Deep residual learning for image recognition

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 770–778, 2016

    2016

  5. [5]

    Mingxing Tan and Quoc V. Le. Efficientnet: Rethink- ing model scaling for convolutional neural networks. In Proceedings of the 36th International Conference on Machine Learning (ICML), volume 97 ofProceed- ings of Machine Learning Research, pages 6105–6114, 2019

    work page 2019

  6. [6]

    Improving un- supervised defect segmentation by applying struc- tural similarity to autoencoders

    Paul Bergmann, Sindy Löwe, Michael Fauser, David Sattlegger, and Carsten Steger. Improving un- supervised defect segmentation by applying struc- tural similarity to autoencoders. InarXiv preprint arXiv:1807.02011, 2018. URL https://arxiv.org/ abs/1807.02011

    arXiv 2018

  7. [7]

    Memorizing normality to de- tect anomaly: Memory-augmented deep autoencoder for unsupervised anomaly detection

    Dong Gong, Lingqiao Liu, Victoria Le, Budhaditya Saha, Moussa Reda Mansour, Svetha Venkatesh, and Anton van den Hengel. Memorizing normality to de- tect anomaly: Memory-augmented deep autoencoder for unsupervised anomaly detection. InProceedings of the IEEE/CVF International Conference on Com- puter Vision (ICCV), pages 1705–1714, 2019. 14

    work page 2019

  8. [8]

    Wald- stein, Ursula Schmidt-Erfurth, and Georg Langs

    Thomas Schlegl, Philipp Seeböck, Sebastian M. Wald- stein, Ursula Schmidt-Erfurth, and Georg Langs. Un- supervised anomaly detection with generative adver- sarial networks to guide marker discovery. InInfor- mation Processing in Medical Imaging (IPMI), pages 146–157, 2017

    work page 2017

Show all 22 references
  1. [9]

    Samet Akçay, Amir Atapour-Abarghouei, and Toby P. Breckon. GANomaly: Semi–supervised anomaly de- tection via adversarial training. InAsian Conference on Computer Vision (ACCV) Workshops, pages 622– 637, 2018

    2018

  2. [10]

    Sub-image anomaly detection with deep pyramid correspondences

    Niv Cohen and Yedid Hoshen. Sub-image anomaly detection with deep pyramid correspondences. InIn- ternational Conference on Learning Representations, 2021

    2021

  3. [11]

    To- wards total recall in industrial anomaly detection

    Karsten Roth, Latha Pemula, Joaquin Zepeda, Bern- hard Schölkopf, Thomas Brox, and Peter Gehler. To- wards total recall in industrial anomaly detection. In Proceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition, pages 14318– 14328, 2022. doi: 10.48...

    2022 doi arXiv

  4. [12]

    Uninformed stu- dents: Student–teacher anomaly detection with dis- criminative latent embeddings

    Paul Bergmann, Sindy Löwe, Michael Fauser, David Sattlegger, and Carsten Steger. Uninformed stu- dents: Student–teacher anomaly detection with dis- criminative latent embeddings. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CV...

    2020

  5. [13]

    Cut- paste: Self-supervised learning for anomaly detection and localization

    Yang Li, Mathieu Salzmann, and Pascal Fua. Cut- paste: Self-supervised learning for anomaly detection and localization. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recog- nition Workshops (CVPRW), pages 461–470, 2021

    2021

  6. [14]

    Fastflow: Un- supervised anomaly detection and localization via 2d normalizing flows

    Tian Yu, Yongzhi Su, Boqing Gong, Lanqing Hong, Jianlong Fu, and Hongsheng Li. Fastflow: Un- supervised anomaly detection and localization via 2d normalizing flows. In Proceedings of the 31st ACM International Conference on Multimedia (ACM MM), pages 1229–1238, 2023. doi: 10.1...

    2023 doi

  7. [15]

    Draem: A discriminative reconstruction autoencoder for weakly-supervised anomaly detection

    Vitja Zavrtanik, Marko Kristan, and Danijel Skocaj. Draem: A discriminative reconstruction autoencoder for weakly-supervised anomaly detection. InProceed- ings of the IEEE/CVF International Conference on Computer Vision (ICCV), pages 833–842, 2021

    2021

  8. [16]

    Vander- meulen, and et al

    Lukas Ruff, Jacob Meadowcroft, Robert A. Vander- meulen, and et al. Efficient one-class density estima- tion with support vector data description and normal- izing flows. InInternational Conference on Learning Representations (ICLR), 2021. arXiv:2106.09694

    2021 arXiv

  9. [17]

    Light padim for unsupervised defect detection and location

    Chaoqiang Ren and Dengfeng Liu. Light padim for unsupervised defect detection and location. InPro- ceedings of the 8th International Conference on Com- puting and Artificial Intelligence (ICCAI), pages 739– 744, 2022. doi: 10.1145/3532213.3532326

    2022

  10. [18]

    Angelina Ibarra and Joshua Peeples. Padim-ace: Patch distribution modeling framework with adap- tive cosine estimator for anomaly detection and lo- calization in synthetic aperture radar imagery.arXiv preprint arXiv:2504.08049, 2025

    2025 arXiv

  11. [19]

    Group invariant scattering.Com- munications on Pure and Applied Mathematics, 65 (10):1331–1398, 2012

    Stéphane Mallat. Group invariant scattering.Com- munications on Pure and Applied Mathematics, 65 (10):1331–1398, 2012

    2012

  12. [20]

    Anomaly detection in time series data using a combination of wavelets, neural networks and hilbert transform

    Stratis Kanarachos, Jino Mathew, Alexander Chro- neos, and M Fitzpatrick. Anomaly detection in time series data using a combination of wavelets, neural networks and hilbert transform. In 2015 6th In- ternational Conference on Information, Intelligence, Systems and Applications...

    2015

  13. [21]

    De-noising by soft-thresholding

    David L Donoho. De-noising by soft-thresholding. IEEE Transactions on Information Theory, 41(3): 613–627, 1995. doi: 10.1109/18.382009

    1995 doi

  14. [22]

    Pytorch: An imperative style, high-performance deep learning li- brary

    Adam Paszke, Sam Gross, Francisco Massa, Adam Lerer, James Bradbury, Gregory Chanan, Trevor Killeen, Zeming Lin, Natalia Gimelshein, Luca Antiga, Alban Desmaison, Andreas Kopf, Edward Yang, Zachary DeVito, Martin Raison, Alykhan Te- jani, Sasank Chilamkurthy, Benoit Steiner, L...

    2019

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.