Pith. sign in

REVIEW 5 major objections 5 minor 1 cited by

Are machine learning interpretations reliable? A stability study on global interpretations

T0 review · 5 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

Pith's one-line read Popular machine learning interpretation methods are frequently less stable than the predictions they explain, and prediction accuracy does not predict interpretation stability.

desk verdict A broad, useful stability benchmark for global IML interpretations with a plausible core finding, but uncalibrated noise and post-hoc dataset selection weaken the generality. read the letter →

arxiv 2505.15728 v1 pith:JOYBISVH submitted 2025-05-21 stat.ML cs.LGstat.AP

classification stat.MLcs.LGstat.AP
keywords interpretablemachinelearninginterpretationstabilityfeatureimportanceclusteringdimensionreductionperturbationanalysistabularbenchmarkdatasetsIMLreliability
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

This paper asks whether the interpretations produced by interpretable machine learning methods can be trusted, and argues that stability under small data perturbations is a necessary prerequisite. Across 25 tabular benchmark datasets and four task families, the authors measure how much feature importance rankings, clusterings, and low-dimensional embeddings change under random 70/30 train/test splits and additive Gaussian or Laplace noise. They find that interpretations are frequently unstable, that they are typically less stable than the predictions themselves, and that there is no reliable association between prediction accuracy and interpretation stability. The paper also finds that no single interpretation method consistently ranks as most stable across datasets. The implication is that interpretability alone does not warrant trust, and that stability needs to be measured and reported separately from predictive performance.

What carries the argument

The load-bearing machinery is a perturbation-and-resampling protocol paired with a family of stability metrics. Interpretations are recomputed across 100 random 70/30 train/test splits for supervised tasks and across subsampling plus additive Gaussian or Laplace noise (variance up to 5) for unsupervised tasks. Stability is scored with top-K rank metrics (Jaccard similarity, average overlap, and top-K Kendall's tau), clustering agreement metrics (adjusted Rand index, Fowlkes-Mallows index, mutual information, V-measure), and a new nearest-neighbor Jaccard area-under-the-curve score for dimension reduction. These scores quantify how much an interpretation changes under small perturbations, which is the paper's operational definition of reliability.

What would settle it

Run the same protocol on datasets using alternative perturbation types not covered in the paper, such as bootstrap resampling, feature permutation, time-based or grouped splits, and noise scaled to each feature's standard deviation, then check whether top-K Jaccard and average-overlap scores remain low and whether accuracy-stability regression slopes stay flat. If many methods show high stability under realistic perturbations, or if prediction accuracy strongly predicts interpretation stability, the paper's central claim would be contradicted.

Watch

Extended reading notes

Core claim

The paper's central empirical discovery is that stability and accuracy come apart: models with similar prediction accuracy can have very different interpretation stability, and more accurate models do not generally give more stable feature importance rankings. Within-method stability scores for feature importance are often low, while prediction stability is comparatively high; between-method agreement on interpretations is also weak, even when predictions agree strongly. In clustering and dimension reduction, stability is higher on average but still varies sharply by method and dataset, and methods that give stable clustering labels can still give unstable nearest-neighbor structure. The authors conclude that interpretation reliability must be assessed directly, not inferred from predictive performance, and they provide an open-source dashboard and Python package for doing so.

Load-bearing premise

The load-bearing premise is that random 70/30 train/test splits and additive Gaussian or Laplace noise with variance up to 5 constitute 'small random perturbations' that mimic realistic data variability; if real-world distribution shifts behave differently, the measured instability could be overstated or understated.

Editorial extensions

If this is right

  • Practitioners who report a model's accuracy without reporting interpretation stability are missing a key axis of trustworthiness, since high accuracy can coexist with volatile explanations.
  • The choice of an interpretation method should be justified empirically on the specific dataset, because no method is consistently most stable across datasets.
  • Stable predictions do not guarantee stable interpretations, so stability analyses must target the interpretation itself rather than the model output.
  • Unsupervised interpretations, such as cluster assignments or embeddings, should undergo the same stability checks as supervised feature importance before being used for scientific discovery.
  • The released software enables researchers to run the same perturbation-and-metric protocol on their own data and IML methods, making stability assessment a routine step.

Reading between the lines

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

  • An implication the authors leave implicit is that stability scores could be used to build more trustworthy interpretations by identifying which methods agree on a given dataset and focusing on their consensus, rather than trusting a single method's output.
  • The observed instability may extend beyond tabular feature importance and clusterings to other explanation formats such as rules, textual explanations, and visual saliency maps, since the underlying sensitivity to data perturbations is a general phenomenon.
  • A testable extension would be to check whether stability rankings change when perturbations are scaled to each feature's natural variability instead of using a global noise variance, which could reveal whether the reported instability is an artifact of noise levels that are large relative to signal.
  • The accuracy-stability decoupling suggests that benchmarks for interpretable machine learning should treat interpretation stability as a reporting requirement rather than an optional diagnostic, especially in high-stakes applications.
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

5 major / 5 minor

Summary. The paper reports a large-scale empirical study of the stability of global interpretable machine learning (IML) outputs for classification, regression, clustering, and dimensionality reduction on tabular benchmark data. Stability is operationalized as agreement under repeated 70/30 train/test splits and under additive Gaussian or Laplace noise, measured with Jaccard, average overlap, top-K Kendall's tau, ARI, Fowlkes-Mallows, mutual information, V-measure, and a proposed NN-Jaccard-AUC metric. Across roughly 50 IML method-dataset combinations, the authors report that interpretations are frequently unstable, are notably less stable than the predictions themselves, show no association with predictive accuracy, and that no single method is most stable across datasets. The paper also releases an interactive dashboard and a Python package for users to run similar reliability assessments.

Significance. If the findings hold, this is a useful systematic benchmark for the IML community: it goes beyond single-method studies, covers supervised and unsupervised tasks, uses multiple stability metrics, and ships open-source tooling for practitioners. The scale of the experiment (hundreds of thousands of repeats) and the public dashboard are concrete strengths. However, the abstract-level claims currently outrun the methodological rigor of the study: the perturbation scale is uncalibrated, the dataset inventory is internally inconsistent, some dataset and hyperparameter choices are made after looking at performance, and the headline comparisons with prediction stability are not formalized with uncertainty quantification. These issues do not invalidate the empirical enterprise, but they must be fixed before the broad claims can be accepted as stated.

major comments (5)
  1. [Appendix A.1 and Figure 8] The noise perturbation is not calibrated to feature scale and the parameterization is ambiguous. Appendix A.1 defines noise as Normal(0, sigma^2) or Laplace(0, sigma^2) with variance sigma^2 in [0, 5], while Figure 8 and Section B.2 use 'normal noise with a standard deviation of 0.15' and 'sigma = 0.15'. If sigma is a standard deviation there, the variance is 0.0225, which is not clearly connected to the grid in A.1; if sigma is read as variance, then the 'standard deviation' statement in B.2 is wrong. More importantly, no standardization is described before adding noise, so a fixed magnitude is applied to datasets with very different feature scales (e.g., PANCAN RNA-seq counts versus Iris physical measurements). A fixed noise level can be negligible for high-variance features and overwhelming for low-variance features, which directly affects the NN-Jaccard-AUC stability results in Figure 8 and the claim that dimension-reduction interpretations are 'frequently unstable'. Please state one convention, standardize features or calibrate noise to per-feature variance, and report a sensitivity sweep over noise levels.
  2. [Section 2.4 and Table 5] The dataset inventory is internally inconsistent. Table 5 lists 25 datasets, but Section 2.4 and the results refer to datasets that do not appear in the table, including DNase, Madelon, Amphibians, Theorem, Digit MNIST, and Darmanis. Appendix C states that the study covers 'over 50 IML methodologies, 33 data sets' and that the classification/regression pages use 13 datasets, neither of which matches Table 5. This mismatch prevents the reader from reconstructing the actual experimental scope and undermines the reproducibility of the paper's central claims. Please provide a complete dataset table and reconcile the counts throughout the text and appendix.
  3. [Section 2.4] The dataset inclusion rules appear to be applied after seeing the methods' performance. The text says 'We select classification data sets whose ARIs are all above 0.2' and, for dimension reduction, excludes datasets whose clustering accuracy on the first two reduced dimensions falls below an ARI cutoff of 0.2, with additional cutoffs P>15 and N<2000. It is unclear what 'ARIs' means for classification datasets, since ARI is defined as a clustering similarity metric and not a classification accuracy measure. If these thresholds were fixed after inspecting results, the benchmark is restricted to datasets on which the studied methods achieve a minimum level of accuracy, which can only weaken the accuracy-stability association and bias the 'no single method' conclusion. Please state when and why the thresholds were chosen, justify them independently of the methods' performance, and report results before and after the exclusions.
  4. [Section 2.2.3 and Table 3] The hierarchical clustering distance metric is selected per dataset by the highest clustering accuracy on the true labels. This means the HC stability rankings in Figure 4 are not the stability of a fixed, pre-specified method; the selection itself uses label information on the same datasets whose stability is later compared. The choice among distance metrics is a hyperparameter and should be made by a procedure that does not leak the test labels, or results for all distances should be reported so the reader can assess the sensitivity of the stability rankings.
  5. [Section 3.2 and abstract] The claim that interpretations are 'notably less stable than the predictions themselves' is supported only by visual comparison of heatmaps (A) and (E) in Figures 2 and 3, without a paired statistical test or an effect-size summary across datasets and methods. Similarly, the 'no single method consistently provides the most stable interpretations' claim rests on bump plots (Figures 2B-5B) that rank methods without error bars or confidence intervals; with 50-100 repeats per setting, the rank differences may be within sampling variation. Please add uncertainty quantification for the stability scores and a formal comparison of interpretation stability versus prediction stability, or temper the abstract-level wording accordingly.
minor comments (5)
  1. [Equation (2)] The definition of top-K Kendall's tau in Eq. (2) uses the term \bar{K}^{(p)}_{i,j} without giving its piecewise definition; please include the missing definition from the cited Fagin et al. reference.
  2. [Throughout] There are repeated typos and grammatical slips, including 'spannining' (Section 2), 'are have' (Section 3.2), 'logiaclly' (Section 3.3), 'underlyign' (Discussion), 'paramter' (Appendix A.3.1), 'featuers' (Appendix C.2.1), 'Tetragonla' (Appendix B.1), and 'Figure 4 P-values' in the Appendix B.3 caption; these should be corrected.
  3. [Section 2.2.2 and Table 2] The Deep Autoencoder row in Table 2 uses em-dashes for both the Global/Local and Linear/Non-linear columns; please specify the intended categorization or explain why it does not apply.
  4. [Appendix A.2.1] The NN-Jaccard-AUC metric is defined in prose but the formula in Table 4 is a compact notation; please state explicitly that it is an average over samples and over the 50 selected values of K, as described in the appendix, so the table is self-contained.
  5. [References] Several references are incomplete or garbled, notably the author string in reference [3]; please verify all references against the publisher records.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the stability measurements are empirical and self-contained; the reliability-stability link is an explicit definitional stance, not a hidden derivation.

full rationale

The paper's central claims—that popular interpretation methods are frequently unstable, that interpretation stability is lower than prediction stability, that accuracy is not associated with stability, and that no method is consistently most stable—are empirical results obtained by applying defined stability metrics (Jaccard, AO, Kendall's Tau, ARI, NN-Jaccard-AUC) to interpretations across repeated train/test splits and noise perturbations. Nothing in these measurements is fitted to the conclusions, and no fitted parameter is renamed as a prediction. The one terminological loop is explicit: Section 1 defines stability as findings that are 'consistent or reliable under small random perturbations,' and Appendix A.1 states 'We define an interpretation to be reliable if the same or similar interpretations can be derived from new data of the same distribution. Therefore, the reliability of a machine learning model can be measured by the stability of its derived interpretations.' The conclusion that unstable interpretations are unreliable follows from that definitional stance rather than from a hidden equivalence; the empirical content lies in the measured instability itself, which is not derived from the conclusion. Self-citations (Allen, Gan, Zheng 2023; Gan, Zheng, Allen 2022; Zikry and Allen 2025) appear only as background references and are not load-bearing for the stability measurements. The choice of perturbation scale (e.g., sigma^2 in [0,5]) and the mixing of variance versus standard deviation parameterizations are legitimate correctness concerns, but they are not circularity. The paper is self-contained against external benchmarks, so the score is 0.

Assumptions & free parameters 4 free parameters · 3 assumptions · 1 invented entities

The central claim rests on design choices (K, noise level, dataset cutoffs, per-dataset hyperparameter selection) rather than fitted parameters in a model. These choices are reasonable but arbitrary, and they affect the reported stability scores. The domain assumptions about perturbations and stability-as-reliability are load-bearing.

free parameters (4)
  • Top-K features (K=10) = 10
    Default number of top features used for stability metrics in the main figures; stability scores vary with K, as shown in Appendix C.2.1.
  • Noise level for dimension reduction (sigma) = 0.15
    Standard deviation of Gaussian noise added for the NN-Jaccard-AUC analysis in Figure 8; different noise levels would likely change stability scores.
  • Dataset inclusion cutoffs = ARI>0.2, P>15, N<2000
    Section 2.4 excludes datasets based on clustering accuracy, number of features, and size for dimension reduction tasks, which may bias the dataset composition.
  • Per-dataset distance metric for hierarchical clustering = varies (Table 3)
    The distance function with the highest clustering accuracy is selected for each dataset, which is a form of fitting to the data and may affect measured stability.
assumptions (3)
  • domain assumption Stability is a prerequisite for reliability
    Section 1 states that stability is a broad prerequisite criterion to trust interpretations; this is assumed, not proven, and underpins the interpretation of the results.
  • domain assumption Random train/test splits and additive noise represent small perturbations
    Section 2.3/A.1 assume these perturbations capture sensitivity to realistic data variations; if the perturbations are not appropriate, the stability measurements lose meaning.
  • domain assumption Standard implementations of ML methods are unbiased representations
    The paper relies on default sklearn and DeepExplain implementations without verifying that they reflect the methods as used in practice.
invented entities (1)
  • NN-Jaccard-AUC metric
    purpose: Measures local neighbor stability in dimension reduction by computing the AUC of Jaccard similarity of nearest neighbors against neighbor count.
    Introduced in Section 2.3 to quantify visualization reliability; it has no external validation or benchmark and is only used within this study.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Are machine learning interpretations reliable? A stability study on global interpretations." pith.science (2026). https://pith.science/paper/JOYBISVH

@misc{pith2026250515728,
  author       = {Pith},
  title        = {Pith review of: Are machine learning interpretations reliable? A stability study on global interpretations},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/JOYBISVH}},
  note         = {Machine review of arXiv:2505.15728}
}
read the original abstract

As machine learning systems are increasingly used in high-stakes domains, there is a growing emphasis placed on making them interpretable to improve trust in these systems. In response, a range of interpretable machine learning (IML) methods have been developed to generate human-understandable insights into otherwise black box models. With these methods, a fundamental question arises: Are these interpretations reliable? Unlike with prediction accuracy or other evaluation metrics for supervised models, the proximity to the true interpretation is difficult to define. Instead, we ask a closely related question that we argue is a prerequisite for reliability: Are these interpretations stable? We define stability as findings that are consistent or reliable under small random perturbations to the data or algorithms. In this study, we conduct the first systematic, large-scale empirical stability study on popular machine learning global interpretations for both supervised and unsupervised tasks on tabular data. Our findings reveal that popular interpretation methods are frequently unstable, notably less stable than the predictions themselves, and that there is no association between the accuracy of machine learning predictions and the stability of their associated interpretations. Moreover, we show that no single method consistently provides the most stable interpretations across a range of benchmark datasets. Overall, these results suggest that interpretability alone does not warrant trust, and underscores the need for rigorous evaluation of interpretation stability in future work. To support these principles, we have developed and released an open source IML dashboard and Python package to enable researchers to assess the stability and reliability of their own data-driven interpretations and discoveries.

Figures

Figures reproduced from arXiv: 2505.15728 by the authors.

Figure 1
Figure 1. Overview of study design with data splitting. After conducting multiple random data splitting, [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. IML Performance on Classification Tasks. A: Heatmap of within-method interpretation stability. B: Bump plot of IML methods ranked by the level of interpretation stability. C: Heatmap of between-method interpretation stability. D: Heatmap of between-method prediction accuracy on test sets. E: Heatmap of prediction stability on test sets. F: Scatterplot of accuracy and interpretation stability, colored by data sets, w… view at source ↗
Figure 3
Figure 3. IML Performance on Regression Tasks. A: Heatmap of within-method interpretation stability. B: Bump plot of IML methods ranked by the level of interpretation stability. C: Heatmap of between-method interpretation stability. D: Heatmap of between-method prediction accuracy on test sets. E: Heatmap of prediction stability on test sets. F: Scatterplot of accuracy and interpretation stability, colored by data sets, with … view at source ↗
Figures from the paper (19 more)
Figure 4
Figure 4. Figure 4: IML Performance on Clustering Methods. A: Heatmap of within-method interpretation stability. B: Bump plot of IML methods ranked by the level of interpretation stability. C: Heatmap of between-method interpretation stability. D: Heatmap of between-method prediction accu…
Figure 8
Figure 8. Figure 8: Comparing the within-method stability heatmaps (A) in Figure 5 and Figure 8 and bump plots [PITH_FULL_IMAGE:figures/full_fig_p014_8.png]
Figure 5
Figure 5. Figure 5: IML Performance on Dimension Reduction Methods. A: Heatmap of within-method interpretation stability. B: Bump plot of IML methods ranked by the level of interpretation stability. C: Heatmap of between-method interpretation stability. D: Heatmap of between-method predic…
Figure 6
Figure 6. Figure 6: One may ask why not use the AO or Kendall’s Tau, which are used as metrics in the feature [PITH_FULL_IMAGE:figures/full_fig_p021_6.png]
Figure 7
Figure 7. Figure 7: Scatterplot of the exponential of negative MSE and MSE, colored by machine learning models of [PITH_FULL_IMAGE:figures/full_fig_p033_7.png]
Figure 8
Figure 8. Figure 8: Comparing the within-method stability heatmaps (A) in Figure 5 and Figure 8 and bump plots [PITH_FULL_IMAGE:figures/full_fig_p034_8.png]
Figure 8
Figure 8. Figure 8: Interpretation stability of nearest neighbors in dimension reduction IML methods with rank [PITH_FULL_IMAGE:figures/full_fig_p035_8.png]
Figure 9
Figure 9. Figure 9: Detailed Results: Between-method stability Heatmap of Each Data Set. [PITH_FULL_IMAGE:figures/full_fig_p036_9.png]
Figure 10
Figure 10. Figure 10: A: Summary scatterplots for each data in classification. Left: Interpretations stability against prediction accuracy; Middle: Interpretations stability against prediction stability; Right: prediction stability against prediction accuracy. B: P-values of fitted coeffic…
Figure 11
Figure 11. Figure 11: Answering Q3 From the Dashboard. A: Summary scatterplots for each dataset used in regression. Left: Interpretations stability against prediction accuracy; Middle: Interpretations stability against prediction stability; Right: prediction stability against prediction ac…
Figure 12
Figure 12. Figure 12: Dashboard navigation bar. The home page gives an overview of the dashboard, introducing details of the IML tasks, reliability tests, stability metrics, and data sets. The instruction page walks through the usage of the dashboard, including the layout of results pages,…
Figure 13
Figure 13. Figure 13: Screenshot of a list components of in feature importance (classification) page of the dashboard. [PITH_FULL_IMAGE:figures/full_fig_p041_13.png]
Figure 14
Figure 14. Figure 14: Example of Preloaded Datasets Within the Dashboard. [PITH_FULL_IMAGE:figures/full_fig_p042_14.png]
Figure 15
Figure 15. Figure 15: Summary Figure: Stability Heatmap of IML methods in Classification [PITH_FULL_IMAGE:figures/full_fig_p043_15.png]
Figure 16
Figure 16. Figure 16: Summary Figure: Stability Line Plot Aggregated over Datasets) [PITH_FULL_IMAGE:figures/full_fig_p044_16.png]
Figure 17
Figure 17. Figure 17: Summary Figure: Stability Bump Plot of Classification [PITH_FULL_IMAGE:figures/full_fig_p044_17.png]
Figure 18
Figure 18. Figure 18: Detailed Results: Relationship between Interpretation stability and the Number of Top Features [PITH_FULL_IMAGE:figures/full_fig_p045_18.png]
Figure 19
Figure 19. Figure 19: Summary Figure: Between-method stability Heatmap of Classification [PITH_FULL_IMAGE:figures/full_fig_p046_19.png]
Figure 20
Figure 20. Figure 20: Detailed Results: Relationship between Interpretation stability and Predictive Accuracy of [PITH_FULL_IMAGE:figures/full_fig_p047_20.png]

Discussion (0). Sign in to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Unsupervised Machine Learning for Scientific Discovery: Workflow and Best Practices

    cs.LG 2025-06 conditional novelty 4.0 of 10

    A best-practices workflow for unsupervised scientific discovery, illustrated by a stability- and generalizability-driven clustering case study of Milky Way globular clusters using APOGEE data.

Reference graph

Works this paper leans on

151 extracted references · 52 canonical work pages · cited by 1 Pith paper

  1. [1]

    [Accessed 20-05-2025]

    GitHub - kundajelab/deeplift: Public facing deeplift repo — github.com.https://github.com/ kundajelab/deeplift. [Accessed 20-05-2025]

  2. [2]

    DeepExplain also includes support for Shapley Values sampling

    GitHub - marcoancona/DeepExplain: A unified framework of perturbation and gradient-based attribution methods for Deep Neural Networks interpretability. DeepExplain also includes support for Shapley Values sampling. (ICLR 2018) — github.com. https://github.com/marcoancona/ DeepExplain. [Accessed 20-05-2025]

  3. [3]

    B. . W. H. . H. M. S. C. L. . . P. P. J. . K. R. 13, G. data analysis: Baylor College of Medicine Creighton Chad J. 22 23 Donehower Lawrence A. 22 23 24 25, I. for Systems Biology Reynolds Sheila 31 Kreisberg Richard B. 31 Bernard Brady 31 Bressler Ryan 31 Erkkila Timo 32 Lin Jake 31 Thorsson Vesteinn 31 Zhang Wei 33 Shmulevich Ilya 31, et al. Comprehensi...

  4. [4]

    Abdi and L

    H. Abdi and L. J. Williams. Principal component analysis.Wiley interdisciplinary reviews: compu- tational statistics, 2(4):433–459, 2010

  5. [5]

    Achilles, H

    C. Achilles, H. P. Bain, F. Bellott, J. Boyd-Zaharias, J. Finn, J. Folger, J. Johnston, and E. Word. Tennessee’s Student Teacher Achievement Ratio (STAR) project, 2008

  6. [6]

    Adadi and M

    A. Adadi and M. Berrada. Peeking inside the black-box: a survey on explainable artificial intelligence (xai). IEEE access, 6:52138–52160, 2018

  7. [7]

    Local Explanation Methods for Deep Neural Networks Lack Sensitivity to Parameter Values

    J. Adebayo, J. Gilmer, I. Goodfellow, and B. Kim. Local explanation methods for deep neural networks lack sensitivity to parameter values.arXiv preprint arXiv:1810.03307, 2018

  8. [8]

    Adebayo, J

    J. Adebayo, J. Gilmer, M. Muelly, I. Goodfellow, M. Hardt, and B. Kim. Sanity checks for saliency maps. arXiv preprint arXiv:1810.03292, 2018

Show all 151 references
  1. [9]

    Agarwal, J

    S. Agarwal, J. Wills, L. Cayton, G. Lanckriet, D. Kriegman, and S. Belongie. Generalized non-metric multidimensional scaling. InArtificial intelligence and statistics, pages 11–18. PMLR, 2007

  2. [10]

    G. I. Allen, L. Gan, and L. Zheng. Interpretable machine learning for discovery: Statistical challenges and opportunities. Annual Review of Statistics and Its Application, 11, 2023

  3. [11]

    Alomari and M

    Y. Alomari and M. Andó. Shap-based insights for aerospace phm: Temporal feature importance, dependencies, robustness, and interaction analysis.Results in Engineering, 21:101834, 2024

  4. [12]

    Ancona, E

    M. Ancona, E. Ceolini, C. Öztireli, and M. Gross. Towards better understanding of gradient-based attribution methods for deep neural networks.arXiv preprint arXiv:1711.06104, 2017

  5. [13]

    Arthur and S

    D. Arthur and S. Vassilvitskii. k-means++: The advantages of careful seeding. Technical report, Stanford, 2006

  6. [14]

    S. Bach, A. Binder, G. Montavon, F. Klauschen, K.-R. Müller, and W. Samek. On pixel-wise explanations for non-linear classifier decisions by layer-wise relevance propagation. PloS one, 10(7):e0130140, 2015. 48

  7. [15]

    Baehrens, T

    D. Baehrens, T. Schroeter, S. Harmeling, M. Kawanabe, K. Hansen, and K.-R. Müller. How to explain individual classification decisions.The Journal of Machine Learning Research, 11:1803–1831, 2010

  8. [16]

    Balasubramanian and E

    M. Balasubramanian and E. L. Schwartz. The isomap algorithm and topological stability.Science, 295(5552):7–7, 2002

  9. [17]

    Bastani, C

    O. Bastani, C. Kim, and H. Bastani. Interpretability via model extraction. arXiv preprint arXiv:1706.09773, 2017

  10. [18]

    S. Basu, K. Kumbier, J. B. Brown, and B. Yu. Iterative random forests to discover predictive and stable high-order interactions.Proceedings of the National Academy of Sciences, 115(8):1943–1948, 2018

  11. [19]

    R. Berk, L. Brown, A. Buja, K. Zhang, and L. Zhao. Valid post-selection inference.The Annals of Statistics, pages 802–837, 2013

  12. [20]

    C. M. Bishop. Pattern recognition.Machine learning, 128(9), 2006

  13. [21]

    C. Blake. Uci repository of machine learning databases. http://www. ics. uci. edu/˜ mlearn/MLRepository. html, 1998

  14. [22]

    L. Breiman. Random forests.Machine learning, 45(1):5–32, 2001

  15. [23]

    J. P. Bridge, S. B. Holden, and L. C. Paulson. Machine learning for first-order theorem proving. Journal of automated reasoning, 53(2):141–172, 2014

  16. [24]

    Bühlmann, M

    P. Bühlmann, M. Kalisch, and L. Meier. High-dimensional statistics with a view toward applications in biology.Annual Review of Statistics and Its Application, 1:255–278, 2014

  17. [25]

    K. Buza. Feedback prediction for blogs. InData analysis, machine learning and knowledge discovery, pages 145–152. Springer, 2014

  18. [26]

    Campbell, C

    K. Campbell, C. P. Ponting, and C. Webber. Laplacian eigenmaps and principal curves for high resolution pseudotemporal ordering of single-cell rna-seq profiles.bioRxiv, page 027219, 2015

  19. [27]

    Carleo, I

    G. Carleo, I. Cirac, K. Cranmer, L. Daudet, M. Schuld, N. Tishby, L. Vogt-Maranto, and L. Zdeborová. Machine learning and the physical sciences.Reviews of Modern Physics, 91(4):045002, 2019

  20. [28]

    M. A. Carreira-Perpinán. A review of dimension reduction techniques.Department of Computer Science. University of Sheffield. Tech. Rep. CS-96-09, 9:1–69, 1997

  21. [29]

    D. V. Carvalho, E. M. Pereira, and J. S. Cardoso. Machine learning interpretability: A survey on methods and metrics.Electronics, 8(8):832, 2019

  22. [30]

    T. Chen, T. He, M. Benesty, V. Khotilovich, Y. Tang, H. Cho, et al. Xgboost: extreme gradient boosting. R package version 0.4-2, 1(4):1–4, 2015

  23. [31]

    E. P. Consortium et al. An integrated encyclopedia of dna elements in the human genome.Nature, 489(7414):57–74, 2012. 49

  24. [32]

    Cortes and V

    C. Cortes and V. Vapnik. Support-vector networks.Machine learning, 20(3):273–297, 1995

  25. [33]

    Cortez, A

    P. Cortez, A. Cerdeira, F. Almeida, T. Matos, and J. Reis. Modeling wine preferences by data mining from physicochemical properties.Decision support systems, 47(4):547–553, 2009

  26. [34]

    M. L. Davison and S. G. Sireci. Multidimensional scaling. InHandbook of applied multivariate statistics and mathematical modeling, pages 323–352. Elsevier, 2000

  27. [35]

    L. Deng. The mnist database of handwritten digit images for machine learning research [best of the web]. IEEE Signal Processing Magazine, 29(6):141–142, 2012

  28. [36]

    M. F. Dixon, I. Halperin, and P. Bilokon.Machine learning in finance, volume 1170. Springer, 2020

  29. [37]

    Doshi-Velez and B

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

  30. [38]

    Doshi-Velez and B

    F. Doshi-Velez and B. Kim. Considerations for evaluation and generalization in interpretable machine learning. In Explainable and interpretable models in computer vision and machine learning, pages 3–17. Springer, 2018

  31. [39]

    Drozdal, J

    J. Drozdal, J. Weisz, D. Wang, G. Dass, B. Yao, C. Zhao, M. Muller, L. Ju, and H. Su. Trust in automl: exploring information needs for establishing trust in automated machine learning systems. In Proceedings of the 25th International Conference on Intelligent User Interfaces, ...

  32. [40]

    M. Du, N. Liu, and X. Hu. Techniques for interpretable machine learning.Communications of the ACM, 63(1):68–77, 2019

  33. [41]

    Fagin, R

    R. Fagin, R. Kumar, and D. Sivakumar. Comparing top k lists.SIAM Journal on discrete mathematics, 17(1):134–160, 2003

  34. [42]

    Fanaee-T and J

    H. Fanaee-T and J. Gama. Event labeling combining ensemble detectors and background knowledge. Progress in Artificial Intelligence, 2(2):113–127, 2014

  35. [43]

    Fernandes, P

    K. Fernandes, P. Vinagre, and P. Cortez. A proactive intelligent decision support system for predicting the popularity of online news. InPortuguese Conference on Artificial Intelligence, pages 535–546. Springer, 2015

  36. [44]

    R. C. Fong and A. Vedaldi. Interpretable explanations of black boxes by meaningful perturbation. In Proceedings of the IEEE International Conference on Computer Vision, pages 3429–3437, 2017

  37. [45]

    E. B. Fowlkes and C. L. Mallows. A method for comparing two hierarchical clusterings.Journal of the American statistical association, 78(383):553–569, 1983

  38. [46]

    Franck, E

    P. Franck, E. Cameron, G. Good, J.-Y. Rasplus, and B. Oldroyd. Nest architecture and genetic differentiation in a species complex of australian stingless bees.Molecular Ecology, 13(8):2317–2331, 2004. 50

  39. [47]

    L. Gan, L. Zheng, and G. I. Allen. Inference for interpretable machine learning: Fast, model-agnostic confidence intervals for feature importance.arXiv preprint arXiv:2206.02088, 2022

  40. [48]

    L. H. Gilpin, D. Bau, B. Z. Yuan, A. Bajwa, M. Specter, and L. Kagal. Explaining explanations: An overview of interpretability of machine learning. In2018 IEEE 5th International Conference on data science and advanced analytics (DSAA), pages 80–89. IEEE, 2018

  41. [49]

    Goecks, V

    J. Goecks, V. Jalili, L. M. Heiser, and J. W. Gray. How machine learning will transform biomedicine. Cell, 181(1):92–101, 2020

  42. [50]

    J. Gu, Y. Yang, and V. Tresp. Understanding individual decisions of cnns via contrastive backpropa- gation. In Asian Conference on Computer Vision, pages 119–134. Springer, 2018

  43. [51]

    Guidotti, A

    R. Guidotti, A. Monreale, S. Ruggieri, F. Turini, F. Giannotti, and D. Pedreschi. A survey of methods for explaining black box models.ACM computing surveys (CSUR), 51(5):1–42, 2018

  44. [52]

    Guyon, S

    I. Guyon, S. Gunn, A. Ben-Hur, and G. Dror. Result analysis of the nips 2003 feature selection challenge. Advances in neural information processing systems, 17, 2004

  45. [53]

    N. S. Habib, O. K. A. Maghasib, A. R. Al-Ghazali, B. S. Abu-Nasser, and S. S. Abu-Naser. Presence of amphibian species prediction using features obtained from gis and satellite images.International Journal of Academic and Applied Research (IJAAR), 4(11), 2020

  46. [54]

    B. S. Haney. Applied artificial intelligence in modern warfare and national security policy.Hastings Sci. & Tech. LJ, 11:61, 2020

  47. [55]

    Hennig, M

    C. Hennig, M. Meilă, F. Murtagh, and R. Rocci. Handbook of cluster analysis. 2015

  48. [56]

    Huang, Q

    J. Huang, Q. Dong, S. Gong, and X. Zhu. Unsupervised deep learning by neighbourhood discovery. In International Conference on Machine Learning, pages 2849–2858. PMLR, 2019

  49. [57]

    Ignatiev

    A. Ignatiev. Towards trustable explainable ai. InIJCAI, pages 5154–5158, 2020

  50. [58]

    P. T. Inc. Collaborative data science, 2015

  51. [59]

    B. K. Iwana, R. Kuroki, and S. Uchida. Explaining convolutional neural networks using softmax gradient layer-wise relevance propagation. In2019 IEEE/CVF International Conference on Computer Vision Workshop (ICCVW), pages 4176–4185. IEEE, 2019

  52. [60]

    Jung and J

    S. Jung and J. S. Marron. Pca consistency in high dimension, low sample size context.The Annals of Statistics, 37(6B):4104–4130, 2009

  53. [61]

    Kapishnikov, T

    A. Kapishnikov, T. Bolukbasi, F. Viégas, and M. Terry. Xrai: Better attributions through regions. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 4948–4957, 2019

  54. [62]

    M. G. Kendall. A new measure of rank correlation.Biometrika, 30(1/2):81–93, 1938

  55. [63]

    M. A. Kramer. Nonlinear principal component analysis using autoassociative neural networks.AIChE journal, 37(2):233–243, 1991. 51

  56. [64]

    Kraskov, H

    A. Kraskov, H. Stögbauer, and P. Grassberger. Estimating mutual information.Physical review E, 69(6):066138, 2004

  57. [65]

    Kumar and S

    R. Kumar and S. Vassilvitskii. Generalized distances between rankings. InProceedings of the 19th international conference on World wide web, pages 571–580, 2010

  58. [66]

    J. D. Lee, D. L. Sun, Y. Sun, and J. E. Taylor. Exact post-selection inference, with application to the lasso. The Annals of Statistics, 44(3):907–927, 2016

  59. [67]

    Li, C.-Y

    Y. Li, C.-Y. Chen, and W. W. Wasserman. Deep feature selection: theory and application to identify enhancers and promoters.Journal of Computational Biology, 23(5):322–336, 2016

  60. [68]

    Themythosofmodelinterpretability: Inmachinelearning, theconceptofinterpretability is both important and slippery.Queue, 16(3):31–57, 2018

    Z.C.Lipton. Themythosofmodelinterpretability: Inmachinelearning, theconceptofinterpretability is both important and slippery.Queue, 16(3):31–57, 2018

  61. [69]

    Little, L

    C. Little, L. Zheng, and G. Allen. iloco: Distribution-free inference for feature interactions.arXiv preprint arXiv:2502.06661, 2025

  62. [70]

    Liu and M

    B. Liu and M. Udell. Impact of accuracy on model interpretations.arXiv preprint arXiv:2011.09903, 2020

  63. [71]

    H. Liu, K. Roeder, and L. Wasserman. Stability approach to regularization selection (stars) for high dimensional graphical models.Advances in neural information processing systems, 23, 2010

  64. [72]

    E. F. Lock and D. B. Dunson. Bayesian consensus clustering.Bioinformatics, 29(20):2610–2616, 2013

  65. [73]

    Y. Y. Lu, Y. Fan, J. Lv, and W. S. Noble. Deeppink: reproducible feature selection in deep neural networks. arXiv preprint arXiv:1809.01185, 2018

  66. [74]

    Lundberg and S.-I

    S. Lundberg and S.-I. Lee. A unified approach to interpreting model predictions.arXiv preprint arXiv:1705.07874, 2017

  67. [75]

    MacQueen et al

    J. MacQueen et al. Some methods for classification and analysis of multivariate observations. In Proceedings of the fifth Berkeley symposium on mathematical statistics and probability, volume 1, pages 281–297. Oakland, CA, USA, 1967

  68. [76]

    M. M. Martínez-Ortega, L. Delgado, D. C. Albach, J. A. Elena-Rosselló, and E. Rico. Species boundaries and phylogeographic patterns in cryptic taxa inferred from aflp markers: Veronica subgen. pentasepalae (scrophulariaceae) in the western mediterranean.Systematic Botany, 29(4...

  69. [77]

    McInnes, J

    L. McInnes, J. Healy, and J. Melville. Umap: Uniform manifold approximation and projection for dimension reduction. arXiv preprint arXiv:1802.03426, 2018

  70. [78]

    M. Meilă. Comparing clusterings—an information based distance.Journal of multivariate analysis, 98(5):873–895, 2007. 52

  71. [79]

    Meinshausen and P

    N. Meinshausen and P. Bühlmann. Stability selection.Journal of the Royal Statistical Society Series B: Statistical Methodology, 72(4):417–473, 2010

  72. [80]

    Mentch and G

    L. Mentch and G. Hooker. Quantifying uncertainty in random forests via confidence intervals and hypothesis tests. The Journal of Machine Learning Research, 17(1):841–881, 2016

  73. [81]

    E. Min, X. Guo, Q. Liu, G. Zhang, J. Cui, and J. Long. A survey of clustering with deep learning: From the perspective of network architecture.IEEE Access, 6:39501–39514, 2018

  74. [82]

    C. Molnar. Interpretable Machine Learning. 2019

  75. [83]

    Montavon, A

    G. Montavon, A. Binder, S. Lapuschkin, W. Samek, and K.-R. Müller. Layer-wise relevance propagation: an overview. Explainable AI: interpreting, explaining and visualizing deep learning, pages 193–209, 2019

  76. [84]

    Monteleoni, G

    C. Monteleoni, G. A. Schmidt, and S. McQuade. Climate informatics: accelerating discovering in climate science with machine learning.Computing in Science & Engineering, 15(5):32–40, 2013

  77. [85]

    Monti, P

    S. Monti, P. Tamayo, J. Mesirov, and T. Golub. Consensus clustering: a resampling-based method for class discovery and visualization of gene expression microarray data.Machine learning, 52:91–118, 2003

  78. [86]

    C. L. Müller, R. Bonneau, and Z. Kurtz. Generalized stability approach for regularized graphical models. arXiv preprint arXiv:1605.07072, 2016

  79. [87]

    W. J. Murdoch, C. Singh, K. Kumbier, R. Abbasi-Asl, and B. Yu. Definitions, methods, and applications in interpretable machine learning.Proceedings of the National Academy of Sciences, 116(44):22071–22080, 2019

  80. [88]

    W. J. Murdoch, C. Singh, K. Kumbier, R. Abbasi-Asl, and B. Yu. Interpretable machine learning: definitions, methods, and applications.arXiv preprint arXiv:1901.04592, 2019

  81. [89]

    F. Murtagh. A survey of recent advances in hierarchical clustering algorithms.The computer journal, 26(4):354–359, 1983

  82. [90]

    L. H. Nguyen and S. Holmes. Ten quick tips for effective dimensionality reduction.PLoS computational biology, 15(6):e1006907, 2019

  83. [91]

    Papernot and P

    N. Papernot and P. McDaniel. Deep k-nearest neighbors: Towards confident, interpretable and robust deep learning.arXiv preprint arXiv:1803.04765, 2018

  84. [92]

    Pedregosa, G

    F. Pedregosa, G. Varoquaux, A. Gramfort, V. Michel, B. Thirion, O. Grisel, M. Blondel, P. Pretten- hofer, R. Weiss, V. Dubourg, J. Vanderplas, A. Passos, D. Cournapeau, M. Brucher, M. Perrot, and E. Duchesnay. Scikit-learn: Machine learning in Python.Journal of Machine Learnin...

  85. [93]

    D. Pollard. Strong consistency of k-means clustering.The Annals of Statistics, pages 135–140, 1981. 53

  86. [94]

    M. H. Rafiei and H. Adeli. A novel machine learning model for estimation of sale prices of real estate units. Journal of Construction Engineering and Management, 142(2):04015066, 2016

  87. [95]

    W. M. Rand. Objective criteria for the evaluation of clustering methods.Journal of the American Statistical association, 66(336):846–850, 1971

  88. [96]

    Real and J

    R. Real and J. M. Vargas. The probabilistic basis of jaccard’s index of similarity.Systematic biology, 45(3):380–385, 1996

  89. [97]

    D. A. Reynolds et al. Gaussian mixture models.Encyclopedia of biometrics, 741(659-663):3, 2009

  90. [98]

    why should i trust you?

    M. T. Ribeiro, S. Singh, and C. Guestrin. " why should i trust you?" explaining the predictions of any classifier. InProceedings of the 22nd ACM SIGKDD international conference on knowledge discovery and data mining, pages 1135–1144, 2016

  91. [99]

    M. T. Ribeiro, S. Singh, and C. Guestrin. Anchors: High-precision model-agnostic explanations. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 32, 2018

  92. [100]

    L. Rokach. A survey of clustering algorithms.Data mining and knowledge discovery handbook, pages 269–298, 2010

  93. [101]

    J. D. Romano, T. T. Le, W. La Cava, J. T. Gregg, D. J. Goldberg, P. Chakraborty, N. L. Ray, D. Himmelstein, W. Fu, and J. H. Moore. Pmlb v1.0: an open source dataset collection for benchmarking machine learning methods.arXiv preprint arXiv:2012.00058v2, 2021

  94. [102]

    Roscher, B

    R. Roscher, B. Bohn, M. F. Duarte, and J. Garcke. Explainable machine learning for scientific insights and discoveries.IEEE Access, 8:42200–42216, 2020

  95. [103]

    Rosenberg and J

    A. Rosenberg and J. Hirschberg. V-measure: A conditional entropy-based external cluster evaluation measure. In Proceedings of the 2007 joint conference on empirical methods in natural language processing and computational natural language learning (EMNLP-CoNLL), pages 410–420, 2007

  96. [104]

    F. Rossi. Building trust in artificial intelligence.Journal of international affairs, 72(1):127–134, 2018

  97. [105]

    C. Rudin. Stop explaining black box machine learning models for high stakes decisions and use interpretable models instead.Nature Machine Intelligence, 1(5):206–215, 2019

  98. [106]

    Rudin, C

    C. Rudin, C. Chen, Z. Chen, H. Huang, L. Semenova, and C. Zhong. Interpretable machine learning: Fundamental principles and 10 grand challenges.Statistics Surveys, 16:1–85, 2022

  99. [107]

    Sah and E

    P. Sah and E. Fokoué. What do asian religions have in common? an unsupervised text analytics exploration. arXiv preprint arXiv:1912.10847, 2019

  100. [108]

    D. Sculley. Web-scale k-means clustering. InProceedings of the 19th international conference on World wide web, pages 1177–1178, 2010

  101. [109]

    R. R. Selvaraju, A. Das, R. Vedantam, M. Cogswell, D. Parikh, and D. Batra. Grad-cam: Why did you say that?arXiv preprint arXiv:1611.07450, 2016. 54

  102. [110]

    Shi and J

    J. Shi and J. Malik. Normalized cuts and image segmentation.IEEE Transactions on pattern analysis and machine intelligence, 22(8):888–905, 2000

  103. [111]

    G. S. Shieh. A weighted kendall’s tau statistic.Statistics & probability letters, 39(1):17–24, 1998

  104. [112]

    Shrikumar, P

    A. Shrikumar, P. Greenside, and A. Kundaje. Learning important features through propagating activation differences. InInternational Conference on Machine Learning, pages 3145–3153. PMLR, 2017

  105. [113]

    Siau and W

    K. Siau and W. Wang. Building trust in artificial intelligence, machine learning, and robotics.Cutter business technology journal, 31(2):47–53, 2018

  106. [114]

    Simonyan, A

    K. Simonyan, A. Vedaldi, and A. Zisserman. Deep inside convolutional networks: Visualising image classification models and saliency maps.arXiv preprint arXiv:1312.6034, 2013

  107. [115]

    J. T. Springenberg, A. Dosovitskiy, T. Brox, and M. Riedmiller. Striving for simplicity: The all convolutional net.arXiv preprint arXiv:1412.6806, 2014

  108. [116]

    Strobl, A.-L

    C. Strobl, A.-L. Boulesteix, T. Kneib, T. Augustin, and A. Zeileis. Conditional variable importance for random forests.BMC bioinformatics, 9(1):1–11, 2008

  109. [117]

    Strobl, A.-L

    C. Strobl, A.-L. Boulesteix, A. Zeileis, and T. Hothorn. Bias in random forest variable importance measures: Illustrations, sources and a solution.BMC bioinformatics, 8(1):1–21, 2007

  110. [118]

    Sundararajan, A

    M. Sundararajan, A. Taly, and Q. Yan. Axiomatic attribution for deep networks. InInternational Conference on Machine Learning, pages 3319–3328. PMLR, 2017

  111. [119]

    Tibshirani

    R. Tibshirani. Regression shrinkage and selection via the lasso.Journal of the Royal Statistical Society: Series B (Methodological), 58(1):267–288, 1996

  112. [120]

    R. J. Tibshirani, J. Taylor, R. Lockhart, and R. Tibshirani. Exact post-selection inference for sequential regression procedures.Journal of the American Statistical Association, 111(514):600–620, 2016

  113. [121]

    Tolles and W

    J. Tolles and W. J. Meurer. Logistic regression: relating patient characteristics to outcomes.Jama, 316(5):533–534, 2016

  114. [122]

    Tomsett, D

    R. Tomsett, D. Braines, D. Harborne, A. Preece, and S. Chakraborty. Interpretable to whom? a role- based model for analyzing interpretable machine learning systems.arXiv preprint arXiv:1806.07552, 2018

  115. [123]

    Toreini, M

    E. Toreini, M. Aitken, K. Coopamootoo, K. Elliott, C. G. Zelaya, and A. Van Moorsel. The relationship between trust in ai and trustworthy machine learning technologies. InProceedings of the 2020 conference on fairness, accountability, and transparency, pages 272–283, 2020

  116. [124]

    N. G. Trillos and D. Slepčev. A variational approach to the consistency of spectral clustering.Applied and Computational Harmonic Analysis, 45(2):239–281, 2018. 55

  117. [125]

    Van der Maaten and G

    L. Van der Maaten and G. Hinton. Visualizing data using t-sne.Journal of machine learning research, 9(11), 2008

  118. [126]

    van Zetten, G

    W. van Zetten, G. Ramackers, and H. Hoos. Increasing trust and fairness in machine learning applications within the mortgage industry.Machine Learning with Applications, 10:100406, 2022

  119. [127]

    Verma, V

    S. Verma, V. Boonsanong, M. Hoang, K. E. Hines, J. P. Dickerson, and C. Shah. Counterfactual expla- nations and algorithmic recourses for machine learning: A review.arXiv preprint arXiv:2010.10596, 2020

  120. [128]

    Vilone and L

    G. Vilone and L. Longo. Explainable artificial intelligence: a systematic review.arXiv preprint arXiv:2006.00093, 2020

  121. [129]

    Von Luxburg

    U. Von Luxburg. A tutorial on spectral clustering.Statistics and computing, 17:395–416, 2007

  122. [130]

    Von Luxburg, M

    U. Von Luxburg, M. Belkin, and O. Bousquet. Consistency of spectral clustering.The Annals of Statistics, pages 555–586, 2008

  123. [131]

    Wackerly, W

    D. Wackerly, W. Mendenhall, and R. L. Scheaffer.Mathematical statistics with applications. Cengage Learning, 2014

  124. [132]

    D. S. Watson and L. Floridi. The explanation game: a formal framework for interpretable machine learning. In Ethics, Governance, and Policies in Artificial Intelligence, pages 185–219. Springer, 2021

  125. [133]

    Webber, A

    W. Webber, A. Moffat, and J. Zobel. A similarity measure for indefinite rankings.ACM Transactions on Information Systems (TOIS), 28(4):1–38, 2010

  126. [134]

    J. N. Weinstein, E. A. Collisson, G. B. Mills, K. R. M. Shaw, B. A. Ozenberger, K. Ellrott, I. Shmulevich, C. Sander, and J. M. Stuart. The cancer genome atlas pan-cancer analysis project. Nature genetics, 45(10):1113–1120, 2013

  127. [135]

    B. D. Williamson, P. B. Gilbert, N. R. Simon, and M. Carone. A general framework for inference on algorithm-agnostic variable importance.Journal of the American Statistical Association, (just- accepted):1–38, 2021

  128. [136]

    Wiwie, J

    C. Wiwie, J. Baumbach, and R. Röttger. Comparing the performance of biomedical clustering methods. Nature methods, 12(11):1033–1038, 2015

  129. [137]

    J. Wu, H. Liu, H. Xiong, J. Cao, and J. Chen. K-means-based consensus clustering: A unified view. IEEE transactions on knowledge and data engineering, 27(1):155–169, 2014

  130. [138]

    Xanthopoulos

    P. Xanthopoulos. A review on consensus clustering methods.Optimization in Science and Engineering: In Honor of the 60th Birthday of Panos M. Pardalos, pages 553–566, 2014

  131. [139]

    L. Xia, C. Lee, and J. J. Li. Statistical method scdeed for detecting dubious 2d single-cell embeddings and optimizing t-sne and umap hyperparameters.Nature Communications, 15(1):1753, 2024. 56

  132. [140]

    Xu and D

    R. Xu and D. Wunsch. Survey of clustering algorithms.IEEE Transactions on neural networks, 16(3):645–678, 2005

  133. [141]

    Xu and R

    Y. Xu and R. Goodacre. On splitting training and validation set: a comparative study of cross- validation, bootstrap and systematic sampling for estimating the generalization performance of supervised learning. Journal of analysis and testing, 2(3):249–262, 2018

  134. [142]

    Yilmaz, J

    E. Yilmaz, J. A. Aslam, and S. Robertson. A new rank correlation coefficient for information retrieval. In Proceedings of the 31st annual international ACM SIGIR conference on Research and development in information retrieval, pages 587–594, 2008

  135. [143]

    B. Yu. Veridical data science. InProceedings of the 13th international conference on web search and data mining, pages 4–5, 2020

  136. [144]

    Zantalis, G

    F. Zantalis, G. Koulouras, S. Karabetsos, and D. Kandris. A review of machine learning and iot in smart transportation. Future Internet, 11(4):94, 2019

  137. [145]

    M. D. Zeiler and R. Fergus. Visualizing and understanding convolutional networks. InEuropean conference on computer vision, pages 818–833. Springer, 2014

  138. [146]

    Zhang, R

    T. Zhang, R. Ramakrishnan, and M. Livny. Birch: an efficient data clustering method for very large databases. ACM sigmod record, 25(2):103–114, 1996

  139. [147]

    why should you trust my explanation?

    Y. Zhang, K. Song, Y. Sun, S. Tan, and M. Udell. " why should you trust my explanation?" understanding uncertainty in lime explanations.arXiv preprint arXiv:1904.12991, 2019

  140. [148]

    T. M. Zikry and G. I. Allen. Limeade: Local interpretable manifold explanations for dimension evaluations. InICLR 2025 Workshop on Machine Learning for Genomics Explorations

  141. [149]

    T. M. Zikry, S. C. Wolff, J. S. Ranek, H. M. Davis, A. Naugle, N. Luthra, A. A. Whitman, K. M. Kedziora, W. Stallaert, M. R. Kosorok, et al. Cell cycle plasticity underlies fractional resistance to palbociclib in er+/her2- breast tumor cells.Proceedings of the National Academy...

  142. [150]

    J. R. Zilke, E. L. Mencía, and F. Janssen. Deepred–rule extraction from deep neural networks. In International Conference on Discovery Science, pages 457–473. Springer, 2016

  143. [151]

    Ziyan, M

    U. Ziyan, M. R. Sabuncu, W. E. L. Grimson, and C.-F. Westin. Consistency clustering: a robust algorithm for group-wise registration, segmentation and automatic atlas construction in diffusion mri. International journal of computer vision, 85(3):279–290, 2009. 57

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.