Pith. sign in

REVIEW 6 major objections 4 minor 51 references

Reservoir-enhanced Segment Anything Model for Subsurface Diagnosis

T0 review · 6 major / 4 minor · reviewed 2026-08-16 · deepseek-v4-flash

Pith's one-line read Res-SAM pairs SAM with a wave-dynamics reservoir to find GPR anomalies at F1 above 94 percent.

desk verdict Res-SAM is a sensible integration of SAM with 2D-ESN dynamic features, but the reported >85% generalization claim is not yet verified because beta is unreported and the split is not site-disjoint. read the letter →

arxiv 2504.18802 v1 pith:7MRA6GPI submitted 2025-04-26 eess.IV cs.CVcs.LG

classification eess.IVcs.CVcs.LG
keywords GroundPenetratingRadarGPRB-ScandataReservoirComputingSubsurfaceTargetDetectionSegmentAnythingModelAnomalyEchoStateNetworkInteractivesegmentation
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 tries to establish that underground anomalies in ground-penetrating radar (GPR) scans can be detected and categorized accurately using very little labeled data by combining a general-purpose visual segmenter with a signal-dynamics model. It argues that GPR images are not really images: they are recordings of electromagnetic waves, and the information that matters for anomalies lives in how the wave patterns change both along a single trace and across neighboring traces. Res-SAM first uses the Segment Anything Model with a few click prompts to get a rough candidate region, then fits every local patch with a dual-directional echo state network and compares the resulting dynamic features against a bank of features from normal frames. The paper reports that this two-stage approach reaches an AUC of 0.896 and an F1-score of 95.5% under moderate prompts, and remains at 0.866 AUC and 94.2% F1 when only three positive clicks are allowed. If true, this would make rapid, low-cost subsurface inspection practical in new environments with almost no training data.

What carries the argument

The load-bearing object is the Dual-Directional Echo State Network (2D-ESN), a reservoir-computing model whose hidden state at each point is computed from the current input and the hidden states of the horizontally and vertically preceding points, using two fixed random reservoirs. Fitting a patch by next-point prediction yields readout weights $[W_{out}\,\,a]$ that serve as a compact 'dynamic feature' of that patch. The argument runs on the assumption that normal subsurface conditions produce a tight cluster of such features, so the nearest-neighbor distance from a test patch's feature to the feature-bank determines whether that patch is anomalous; threshold $\beta$ in Eq. (8) decides the cut and the merged patches define the final region.

What would settle it

Take a second GPR survey collected on a different road surface or after a season change, use the original 20 frames as the feature bank with $\beta$ unchanged, and measure F1 on the new data; if the score falls to the level of the SAM-only baseline or below the reported 94%, the normal-feature bank assumption fails. A cheaper check is to report the $\beta$ value used and re-run the 3/0 experiment with $\beta$ chosen by cross-validation on the 20 non-target frames.

Watch

Extended reading notes

Core claim

The central claim is that anomaly detection in GPR B-scan data is best treated as a comparison of local wave dynamics rather than a visual segmentation task, and that a reservoir-computing readout can supply the needed dynamic features. Res-SAM's pipeline is: (1) collect 20 non-target frames, extract patches with a sliding window, fit each with 2D-ESN, and store the readout weights in a feature bank; (2) let SAM propose a candidate region from click prompts; (3) for each point in that region, fit a centered patch with 2D-ESN and compute its anomaly likelihood as the L2 distance to the nearest feature in the bank; (4) merge patches whose likelihood exceeds a threshold beta; (5) refit the final region and cluster the resulting features to label anomaly type. The paper asserts that this consistently outperforms six interactive-segmentation baselines across all tested prompt settings, and that the category clustering reaches 0.91 accuracy, 0.85 ARI, and 0.89 NMI.

Load-bearing premise

The whole framework assumes that the 20 randomly chosen non-target frames capture all the normal wave dynamics that will appear in the remaining data, and that a single fixed distance threshold $\beta$ cleanly separates normal variation from anomalies.

Editorial extensions

If this is right

  • Deployment in a new urban area requires only a handful of normal GPR frames and a few clicks per frame, removing the labeled-anomaly bottleneck that limits deep-learning approaches.
  • Because the same 2D-ESN features are used for both detection and category clustering, the framework can label cavity, crack, looseness, pipeline, and manhole in a single pass without retraining.
  • The method's robustness to prompt reduction (F1 above 94% even at 3/0) implies that operators can rely on it in time-critical surveys where careful prompting is impossible.
  • The reported gains over SimpleClick and ScribblePrompt suggest that wave-dynamics features capture signal that pure visual encoders miss, which could transfer to other wave-based imaging modalities.

Reading between the lines

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

  • Beyond the paper: the nearest-neighbor threshold $\beta$ is described as 'predefined' but its value and selection procedure are not reported; an honest comparison would require stating how $\beta$ is set on the 20 normal frames and whether it is stable across sites.
  • Beyond the paper: the feature bank built from 20 frames is a small sample of normal variability; on roads with changing soil moisture, layering, or antenna coupling, the bank may need periodic refresh, and the paper does not test this drift.
  • Beyond the paper: the same two-stage idea (visual proposer plus reservoir dynamics refiner) could be applied to other non-visual imaging data such as ultrasonic or microwave tomography, where boundaries are also gradual.
  • Beyond the paper: a direct testable extension is to replace the hand-picked threshold $\beta$ with a calibrated quantile of the bank distances, which would make the method parameter-free and easier to reproduce.
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.

Referee Report

6 major / 4 minor

Summary. The paper proposes Res-SAM, a two-phase framework for detecting and categorizing subsurface anomalies in GPR B-scan data. In the first phase, a feature bank is built by fitting local patches from 20 non-target frames with a dual-directional echo state network (2D-ESN) and storing the fitted readout weights as normal dynamic features. In the second phase, SAM with click prompts proposes a candidate anomaly region; each point in that region is represented by a local patch fitted with 2D-ESN, and the resulting feature is scored by its L2 distance to the nearest feature in the bank, with a threshold beta deciding anomaly membership. Identified patches are merged into final rectangular regions, and a second 2D-ESN fitting is followed by clustering to categorize anomalies. Experiments on a 626-frame real-world dataset report AUC up to 0.896 and F1 up to 95.5% across prompt settings from 5/5 to 3/0, outperforming six interactive segmentation baselines, and clustering accuracy up to 0.91.

Significance. The central idea is plausible and practically motivated: Res-SAM does not require anomaly-labeled training data, uses only a small number of non-target frames for initialization, avoids training the reservoir, and combines SAM's visual prompting with wave-dynamics features. The paper also makes its code available, which is a concrete strength. If the empirical claims survive closer evaluation, the method would be a useful resource-efficient tool for GPR-based road inspection. The main weakness is not the method's internal logic but the evaluation protocol: the reported F1 depends on an unreported threshold, the data split may not be site-disjoint, and the comparison omits GPR-specific baselines from the cited literature. These issues are fixable within the scope of the manuscript, so the appropriate outcome is major revision rather than rejection.

major comments (6)
  1. [Section 2.3/Table 1 and Section 4.3.3, Eq. (8)] The anomaly threshold beta is described only as 'predefined', and no value or selection procedure is reported anywhere in the manuscript. The headline F1 numbers in Table 1 (e.g., 95.5% in the 5/5 and 5/3 settings) are computed from the binary classifier in Eq. (8), so these claims cannot be reproduced or checked for test-set threshold tuning. Please report the beta value used for each prompt setting, state whether it was fixed before evaluation, and justify it a priori (for example, as a quantile of the feature-bank distance distribution). Reporting F1 across a range of beta or a precision-recall curve would also clarify how sensitive the result is to this parameter.
  2. [Section 2.2 and Section 2.3] The generalization claim is not supported by the evaluation split. Section 2.2 states that frames cover approximately 15 meters of continuous road B-scan, and Section 2.3 states that 20 non-target frames were 'randomly chosen' from the dataset. If these initialization frames come from the same road segments as the test frames, the feature bank in Eq. (4) can contain near-duplicates of normal patches in the test set; the nearest-neighbor distances in Eq. (7) would then be artificially small for normal patches, inflating the separation exploited in Eq. (8). The paper's claim of applicability across 'diverse environments' requires a site-disjoint or survey-disjoint split, not a random frame-level split. Please also report results over repeated random initializations of the 20-frame bank, with means and standard deviations.
  3. [Section 2.3, Table 1] The evaluation protocol is under-defined. The text says a detection is correct if its IoU with ground truth exceeds 0.5, but it does not specify how AUC and F1 are computed from this rule: are the units frames, candidate regions, or individual patches? How are frames with multiple anomalies handled, and what is the positive/negative definition for the ROC analysis? Without a precise scoring protocol, the numbers in Table 1 are not reproducible. Please specify the evaluation units and the exact procedure used to aggregate detections into AUC and F1.
  4. [Section 2.3] The comparison is limited to interactive segmentation methods, which weakens the 'outperforms state-of-the-art' claim. The sentence 'there is a lack of research specifically focused on anomaly detection in GPR data' is contradicted by the authors' own references, including Refs. [18], [19], [25], and [26], which describe GPR anomaly detection methods. At least one or two recent GPR-specific detection baselines should be included under the same evaluation protocol, or the claim should be narrowed to 'outperforms interactive segmentation baselines'.
  5. [Section 2.4 and Section 4.3.4] The anomaly categorization experiments do not report the cluster count k used for K-Means, Agglomerative Clustering, and Fuzzy C-Means, nor do they describe how cluster labels were matched to ground-truth categories before computing accuracy, ARI, and NMI. Since k is a free parameter and these metrics depend on it, the clustering results in Table 2 cannot be reproduced as reported. Please state the value of k and the label-matching procedure.
  6. [Section 4.1 and Section 4.2] Several central hyperparameters are not reported: patch size X x Y, sliding stride s, reservoir dimensions and spectral radii of Wx and Wy, and the ridge regularization lambda in Eq. (3). These values are needed to reproduce the feature bank and the anomaly detection pipeline. Please provide them in the text or in a supplementary table.
minor comments (4)
  1. [Figures 3 and 6] The manuscript contains raw hexadecimal-looking path fragments after Fig. 3 and Fig. 6; these appear to be formatting corruption and should be removed.
  2. [Table 2] The header for the FCM columns lists 'Acc NMI NMI'; this is likely a typo and should read 'Acc ARI NMI'.
  3. [References] Several references are incomplete: Ref. [20] lacks publication details, Refs. [30] and [36] lack years or page ranges, and Ref. [36] lacks the conference proceedings. Please complete these entries.
  4. [Discussion] The Discussion section's limitation paragraph notes that anomalies not indicated by human prompts are not automatically verified; this scope should be stated more prominently in the abstract or results, since the reported detection accuracy applies to prompted candidate regions rather than fully automatic scanning.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: Res-SAM's anomaly score is a distance to non-target features, not a re-statement of labels; the main open issues are reproducibility of the unreported threshold, not circular derivation.

full rationale

Res-SAM's detection pipeline is self-contained and non-circular. Normal features are obtained by 2D-ESN fitting of non-target patches (Eqs. 1-4), test patches are fitted by the same mechanism, and anomaly likelihood is the L2 distance to the nearest normal feature (Eq. 7) with a threshold decision (Eq. 8). None of these quantities is defined in terms of the anomaly labels or the final IoU/F1 metrics; the labels are used only for evaluation, and the clicked prompts serve as inputs to SAM rather than as fitted outputs of Res-SAM. The categorization stage likewise clusters 2D-ESN features with no label fit. The paper cites earlier model-space learning work by the same group ([7], [19], [25], [26], [51]), but the equations in Section 4 fully specify the 2D-ESN and feature bank, so the argument does not reduce to a self-citation chain. The omission of the numeric value of beta in Eq. (8) and the lack of a site-disjoint split are evaluation and reproducibility concerns, not circularity: beta is called 'predefined', and the nearest-neighbor score is not constructed from the reported metrics. The stated limitation about anomalies not initially indicated by prompts further confirms that the framework does not claim to derive its stated performance from its own assumptions.

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

Res-SAM is an empirical ML pipeline, so the ledger consists mainly of unstated hyperparameters and domain assumptions about GPR wave dynamics. There are no newly postulated physical entities. The most important free parameter is the anomaly threshold beta, because the headline F1 numbers depend on it, and it is never specified.

free parameters (6)
  • Patch size X x Y
    Controls the local GPR context fed into 2D-ESN; never specified numerically (Section 4.2).
  • Sliding stride s
    Determines patch density for the feature bank; not reported (Section 4.2).
  • Reservoir dimensions and spectral radius of Wx, Wy
    Random reservoir weights determine the dynamic features; paper only says spectral radius is between 0 and 1 (Section 4.1).
  • Ridge regularization lambda
    Used to solve output weights in Eq. (3); no value or selection method given.
  • Anomaly threshold beta
    Eq. (8) uses beta to classify patches as normal or anomalous; called 'predefined' but never specified, and it directly controls reported F1.
  • Cluster count k for anomaly categorization
    Clustering in Section 4.3.4 requires a number of clusters; the paper does not state whether k was set to the known six categories or chosen otherwise.
assumptions (4)
  • domain assumption GPR B-scan data can be represented as a 2D signal whose local dual-directional changing information is adequately captured by 2D-ESN next-point prediction.
    Underlies the entire feature extraction; stated in Section 4.1 but not empirically compared to other signal representations.
  • domain assumption Normal features extracted from 20 non-target frames are representative of all normal conditions in the test frames.
    The feature bank in Eq. (4) is built from this small sample; generalization depends on it (Sections 2.3 and 4.2).
  • standard math Echo State Property and random reservoir weights with spectral radius in (0,1) are sufficient for stable 2D reservoir dynamics.
    Taken from reservoir computing literature [48-50]; used to justify the hidden-state iteration in Eq. (1).
  • standard math Ridge regression solution in Eq. (3) gives a well-posed least-squares fit for output weights.
    Standard linear algebra; used without proof.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Reservoir-enhanced Segment Anything Model for Subsurface Diagnosis." pith.science (2026). https://pith.science/paper/7MRA6GPI

@misc{pith2026250418802,
  author       = {Pith},
  title        = {Pith review of: Reservoir-enhanced Segment Anything Model for Subsurface Diagnosis},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/7MRA6GPI}},
  note         = {Machine review of arXiv:2504.18802}
}
read the original abstract

Urban roads and infrastructure, vital to city operations, face growing threats from subsurface anomalies like cracks and cavities. Ground Penetrating Radar (GPR) effectively visualizes underground conditions employing electromagnetic (EM) waves; however, accurate anomaly detection via GPR remains challenging due to limited labeled data, varying subsurface conditions, and indistinct target boundaries. Although visually image-like, GPR data fundamentally represent EM waves, with variations within and between waves critical for identifying anomalies. Addressing these, we propose the Reservoir-enhanced Segment Anything Model (Res-SAM), an innovative framework exploiting both visual discernibility and wave-changing properties of GPR data. Res-SAM initially identifies apparent candidate anomaly regions given minimal prompts, and further refines them by analyzing anomaly-induced changing information within and between EM waves in local GPR data, enabling precise and complete anomaly region extraction and category determination. Real-world experiments demonstrate that Res-SAM achieves high detection accuracy (>85%) and outperforms state-of-the-art. Notably, Res-SAM requires only minimal accessible non-target data, avoids intensive training, and incorporates simple human interaction to enhance reliability. Our research provides a scalable, resource-efficient solution for rapid subsurface anomaly detection across diverse environments, improving urban safety monitoring while reducing manual effort and computational cost.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

51 extracted references · 49 canonical work pages

  1. [18]

    & Chen, H

    Wei, C., Zhou, X., Liu, S. & Chen, H. Enhanced anomaly detection in GPR data by combining spatial and dynamic information.IEEE Transactions on Geoscience and Remote Sensing 62, 1–10 (2024)

  2. [25]

    & Chen, H

    Chen, A., Zhou, X., Fan, Y. & Chen, H. Underground diagnosis based on GPR and learning in the model space. IEEE Transactions on Pattern Analysis and Machine Intelligence 46, 3832–3844 (2024)

  3. [19]

    Zhou, X. et al. Underground anomaly detection in GPR data by learning in the C3 model space. IEEE Transactions on Geoscience and Remote Sensing 1–11 (2023). 18

  4. [26]

    & Chen, H

    Zhou, X., Liu, S., Chen, A. & Chen, H. Learning in CubeRes model space for anomaly detection in 3D GPR data . Proceedings of the 33rd International Joint Conference on Artificial Intelligence (IJCAI-24)

  5. [1]

    Liu, J. et al. Urban sensing using existing fiber-optic networks. Nature Communications 16, 3091 (2025)

  6. [2]

    World cities report: Envisaging the future of cities

    Habitat, U. World cities report: Envisaging the future of cities. United Nations Human Settlements Programme (UN-Habitat), Nairobi, Kenya (2022)

  7. [3]

    Xie, Z. et al. A narrative of urban underground pipeline system disasters in china in 2021: Spatial and temporal distribution, causal analysis, and response strategies. Sustainability 15, 10067 (2023)

  8. [4]

    & Chen, H

    Zhou, X., Chen, Q., Lyu, S. & Chen, H. Mapping the buried cable by ground pen- etrating radar and gaussian-process regression. IEEE Transactions on Geoscience and Remote Sensing 60, 1–12 (2022)

Show all 51 references
  1. [5]

    Shen, H. et al. Quality evaluation of ground improvement by deep cement mixing piles via ground-penetrating radar. Nature Communications 14, 3448 (2023)

  2. [6]

    Conyers, L. B. Ground penetrating radar. Encyclopedia of Imaging Science and Technology (2002)

  3. [7]

    Zhou, X., Chen, H. & Li, J. An automatic GPR B-scan image interpreting model. IEEE Transactions on Geoscience and Remote Sensing 56, 3398–3412 (2018). 17

  4. [8]

    & Chen, H

    Liu, S., Zhou, X. & Chen, H. From data to D3 model: Adaptive subsurface anomaly detection in gpr data. IEEE Transactions on Geoscience and Remote Sensing 62, 1–12 (2024)

  5. [9]

    Lee, K. L. & Mokji, M. M. Automatic target detection in GPR images using histogram of oriented gradients (HOG) . 2014 2nd International Conference on Electronic Design (ICED), 181–186. IEEE

  6. [10]

    Ye, F., Zhang, Z., Chakrabarty, K. & Gu, X. Board-level functional fault diag- nosis using multikernel support vector machines and incremental learning. IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems 33, 279–290 (2014)

  7. [11]

    Todkar, S. S. et al. Detection of debondings with ground penetrating radar using a machine learning method. 2017 9th International Workshop on Advanced Ground Penetrating Radar (IWAGPR), 1–6. IEEE

  8. [12]

    & Khan, U

    Noreen, T. & Khan, U. S. Using pattern recognition with HOG to automatically detect reflection hyperbolas in ground penetrating radar data . 2017 Interna- tional Conference on Electrical and Computing Technologies and Applications (ICECTA), 1–6. IEEE

  9. [13]

    B., Seyfi, L

    Ozkaya, U., Melgani, F., Bejiga, M. B., Seyfi, L. & Donelli, M. GPR B-scan image analysis with deep learning methods. Measurement 165, 107770 (2020)

  10. [14]

    & Wang, J

    Zhu, M., Jiao, L., Liu, F., Yang, S. & Wang, J. Residual spectral–spatial attention network for hyperspectral image classification. IEEE Transactions on Geoscience and Remote Sensing 59, 449–462 (2021)

  11. [15]

    Hou, F., Lei, W., Li, S. & Xi, J. Deep learning-based subsurface target detection from GPR scans. IEEE Sensors Journal 21, 8161–8171 (2021)

  12. [16]

    Liu, B. et al. GPRInvNet: Deep learning-based ground-penetrating radar data inversion for tunnel linings. IEEE Transactions on Geoscience and Remote Sensing 59, 8305–8325 (2021)

  13. [17]

    Liu, Z. et al. Automatic recognition of pavement cracks from combined GPR B-scan and C-scan images using multiscale feature fusion deep neural networks. Automation in Construction 146, 104698 (2023)

  14. [20]

    Kirillov, A. et al. Segment anything . Proceedings of the IEEE/CVF International Conference on Computer Vision , 4015–4026

  15. [21]

    Ma, J. et al. Segment anything in medical images. Nature Communications 15, 654 (2024)

  16. [22]

    Oh, Y. et al. LLM-driven multimodal target volume contouring in radiation oncology. Nature Communications 15, 9186 (2024)

  17. [23]

    Hekrdla, M. et al. Optimized molecule detection in localization microscopy with selected false positive probability. Nature Communications 16, 601 (2025)

  18. [24]

    & Mir-Sanchis, I

    Debiasi-Anders, G., Qiao, C., Salim, A., Li, N. & Mir-Sanchis, I. Phage parasites targeting phage homologous recombinases provide antiviral immunity. Nature Communications 16, 1889 (2025)

  19. [27]

    Maximizing the information return from ground penetrating radar

    Olhoeft, G. Maximizing the information return from ground penetrating radar. Journal of Applied Geophysics 43, 175–187 (2000)

  20. [28]

    & Ralston, J

    Strange, A., Chandran, V. & Ralston, J. Signal processing to improve target detection using ground penetrating radar . 139–143. 4th Australasian Workshop on Signal Processing and Applications

  21. [29]

    & Shan, S

    Pan, T., Tang, L., Wang, X. & Shan, S. Tokenize anything via prompting. arXiv preprint arXiv:2312.09128 (2023)

  22. [30]

    & Niethammer, M

    Liu, Q., Xu, Z., Bertasius, G. & Niethammer, M. Simpleclick: Interactive image segmentation with simple vision transformers . Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , 22290–22300

  23. [31]

    Luo, X. et al. MIDeepSeg: Minimally interactive segmentation of unseen objects from medical images using deep learning. Medical Image Analysis 72, 102102 (2021)

  24. [32]

    E., Rakic, M., Guttag, J

    Wong, H. E., Rakic, M., Guttag, J. & Dalca, A. V. ScribblePrompt: Fast and flexible interactive segmentation for any biomedical image. European Conference on Computer Vision (ECCV) (2024)

  25. [33]

    & Triggs, B

    Dalal, N. & Triggs, B. Histograms of oriented gradients for human detection. 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition 19 (CVPR’05), Vol. 1, 886–893. IEEE

  26. [34]

    & Harwood, D

    Ojala, T., Pietik¨ ainen, M. & Harwood, D. A comparative study of texture mea- sures with classification based on featured distributions. Pattern Recognition 29, 51–59 (1996)

  27. [35]

    & Deng, W

    Liu, S. & Deng, W. Very deep convolutional neural network based image classi- fication using small training sample size . 2015 3rd IAPR Asian Conference on Pattern Recognition (ACPR), 730–734. IEEE

  28. [36]

    & Sun, J

    He, K., Zhang, X., Ren, S. & Sun, J. Deep residual learning for image recog- nition. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 770–778

  29. [37]

    & Wojna, Z

    Szegedy, C., Vanhoucke, V., Ioffe, S., Shlens, J. & Wojna, Z. Rethinking the inception architecture for computer vision . Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , 2818–2826

  30. [38]

    Dosovitskiy, A. et al. An image is worth 16x16 words: Transformers for image recognition at scale. 9th International Conference on Learning Representations (ICLR 2021), Virtual Event, Austria, May 3-7, 2021 . OpenReview.net

  31. [39]

    Deng, J. et al. Imagenet: A large-scale hierarchical image database . 2009 IEEE Conference on Computer Vision and Pattern Recognition , 248–255. Ieee

  32. [40]

    & Murty, M

    Krishna, K. & Murty, M. N. Genetic K-means algorithm. IEEE Transactions on Systems, Man, and Cybernetics, Part B (Cybernetics) 29, 433–439 (1999)

  33. [41]

    Gowda, K. C. & Krishna, G. Agglomerative clustering using the concept of mutual nearest neighbourhood. Pattern Recognition 10, 105–112 (1978)

  34. [42]

    C., Ehrlich, R

    Bezdek, J. C., Ehrlich, R. & Full, W. FCM: The fuzzy c-means clustering algorithm. Computers & Geosciences 10, 191–203 (1984)

  35. [43]

    & Hinton, G

    Van der Maaten, L. & Hinton, G. Visualizing data using t-SNE. Journal of Machine Learning Research 9 (2008)

  36. [44]

    & Wang, J

    Li, D., Han, M. & Wang, J. Chaotic time series prediction based on a novel robust echo state network. IEEE Transactions on Neural Networks and Learning Systems 23, 787–799 (2012)

  37. [45]

    & Yao, X

    Chen, H., Tiˇ no, P. & Yao, X. Cognitive fault diagnosis in Tennessee Eastman Process using learning in the model space. Computers & Chemical Engineering 67, 33–42 (2014)

  38. [46]

    Yan, M. et al. Emerging opportunities and challenges for the future of reservoir computing. Nature Communications 15, 2056 (2024). 20

  39. [47]

    J., Nurdin, H

    Ehlers, P. J., Nurdin, H. I. & Soh, D. Stochastic reservoir computers. Nature Communications 16, 1–11 (2025)

  40. [48]

    echo state

    Jaeger, H. The “echo state” approach to analysing and training recurrent neural networks-with an erratum note. Bonn, Germany: German National Research Center for Information Technology GMD Technical Report 148, 13 (2001)

  41. [49]

    & Young, P

    Buehner, M. & Young, P. A tighter bound for the echo state property. IEEE Transactions on Neural Networks 17, 820–824 (2006)

  42. [50]

    B., Jaeger, H

    Yildiz, I. B., Jaeger, H. & Kiebel, S. J. Re-visiting the echo state property. Neural Networks 35, 1–9 (2012)

  43. [51]

    & Yao, X

    Chen, H., Tiˇ no, P., Rodan, A. & Yao, X. Learning in the model space for cognitive fault diagnosis. IEEE Transactions on Neural Networks and Learning Systems 25, 124–136 (2013). 21

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.