Pith. sign in

REVIEW 4 major objections 5 minor 40 references

Feature Space Topology Control via Hopkins Loss

T0 review · 4 major / 5 minor · reviewed 2026-08-04 · deepseek-v4-flash

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

Pith's one-line read A differentiable Hopkins statistic lets neural networks steer their feature-space topology toward a target arrangement (regular, random, or clustered) with only small changes in task accuracy.

desk verdict The Hopkins loss is a real and useful bias, but the paper's 'enforce topology' claim is not supported by its own numbers. read the letter →

arxiv 2509.11154 v2 pith:KGOFFLNI submitted 2025-09-14 cs.LG cs.AI

Einari Vaaras , Manu Airaksinen This is my paper · ORCID
classification cs.LGcs.AI
keywords HopkinsstatisticfeaturespacetopologylossfunctionregularizationdimensionalityreductionautoencodersclassificationChebyshevdistance
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 introduces Hopkins loss, a differentiable loss function built from the Hopkins statistic, that pushes learned feature representations toward a user-chosen topology: regularly spaced, randomly spaced, or clustered. It matters because controlling feature-space shape could improve generalization, aid dimensionality reduction and visualization, and support transfer learning or robustness to adversarial attacks. The paper demonstrates on speech, text, and image data in both classification and autoencoder settings that adding this loss shifts the feature distribution's H value toward its target while leaving classification accuracy nearly unchanged or even improved. The central claim is that a simple scalar statistic can serve as a practical, low-overhead control knob for feature geometry.

What carries the argument

The Hopkins statistic H, defined as the ratio of nearest-neighbor distances from synthetic uniform points to nearest-neighbor distances from actual sample points, is made differentiable by computing nearest neighbors with hard-min operations (straight-through in the backward pass). The loss LH = |H - HT| then acts as a regularizer that encourages the feature distribution to reproduce a specified H value. Because the authors found Chebyshev distance preserves the intended H ranges across dimensions, they fix D = Chebyshev. The statistic is computed per minibatch, with the synthetic points Y scaled to the minibatch's feature bounding box, and the sample fraction m = 0.05n. This scalar-feedback

What would settle it

A concrete falsifying observation: train a two-dimensional bottleneck autoencoder with HT = 0.01 on any of the paper's datasets and inspect the resulting embedding. The authors' own results show H stays around 0.6–0.7, never entering the claimed regular-spacing range of 0.01–0.3; if the loss truly enforced the desired topology, the embedding would visibly form a periodic grid or regular lattice. More directly: construct two different 2D point sets with identical H values (e.g., a regular grid and a diagonal line) and show that the gradient of LH is identical for both, demonstrating that the sc

Watch

Extended reading notes

Core claim

The paper claims that adding the loss LH = |H - HT| to a classification or autoencoder training objective moves the feature space so that its Hopkins statistic H approaches the target HT. In classification experiments across speech, text, and image data, H moved on average by about 0.09–0.12 toward the target, with classification accuracy essentially preserved (and often improved for text). In bottleneck autoencoder experiments, H moved by about 0.18–0.22, at the cost of a small decrease in downstream linear-classifier accuracy. The loss is computed with Chebyshev distance and with m = 0.05n sampled points, and it adds roughly 10–13% training-time overhead. The authors argue this provides a

Load-bearing premise

The load-bearing premise is that the Hopkins statistic's scalar value (and the published ranges for regular, random, and clustered data) faithfully captures the geometry of the feature space, so that minimizing |H - HT| genuinely controls topology rather than merely moving one nearest-neighbor ratio.

Editorial extensions

If this is right

  • Integrating Hopkins loss into classification training can modify feature topology toward a target H value while keeping classification accuracy effectively unchanged, as shown on speech, text, and image datasets.
  • In dimensionality-reduction autoencoders, the loss produces noticeably larger topology shifts than in classification, enabling low-dimensional embeddings with a desired spacing or clustering structure for visualization or compression.
  • The loss adds only ~10–13% training-time overhead, making it a cheap auxiliary regularizer for large-scale models.
  • The paper's results suggest the method could generalize to other tasks such as transfer learning, generative modeling, and adversarial robustness, where feature topology is believed to matter.
  • The loss is most effective for moving toward clustered or random arrangements; achieving a regularly spaced topology (HT = 0.01) proved difficult, with H staying far above the target range.

Reading between the lines

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

  • Because H is a single scalar, minimizing |H - HT| does not uniquely determine the geometric arrangement; two feature sets with identical H can have very different shapes (e.g., a grid vs. a space-filling curve), so the loss controls a coarse clustering tendency rather than full topology. This is an editorial extension beyond the paper's claims.
  • The choice of Chebyshev distance is load-bearing; the paper notes other metrics (euclidean, cosine) do not preserve the H ranges across dimensions, so the method's portability to new feature spaces likely requires re-validating the distance metric and possibly the H thresholds.
  • One plausible testable extension is combining Hopkins loss with a finer topological descriptor (e.g., persistent homology) to retain the cheap scalar signal while adding sensitivity to global shape; this would likely improve control over regularly spaced arrangements.
  • The observed failure to reach HT = 0.01 suggests the scalar H gradient may be too weak or ambiguous to force a regular grid, implying that deliberate structural regularizers (e.g., repulsive forces or quantization) might be needed to actually achieve regular spacing.
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 Hopkins loss, L_H = |H - H_T|, where H is the Hopkins statistic computed on minibatch feature vectors using Chebyshev distance, and integrates it into MLP training for classification and into bottleneck autoencoder training. The authors claim that minimizing this loss 'enforces' a desired feature-space topology (regularly spaced, randomly spaced, or clustered) across speech, text, and image data. Experiments repeat each configuration 100 times and report classification accuracy and the resulting mean H. The main empirical finding is that adding L_H moves H toward the target relative to baseline in many conditions, with small or moderate changes in classification accuracy. However, the reported H values do not reach the stated target ranges for H_T = 0.01 or H_T = 0.5; only H_T = 0.99 is closely approximated. The authors themselves acknowledge in Section V that L_H 'could not force the features into a regularly-spaced topology,' yet the abstract and introduction continue to use the word 'enforce.'

Significance. If the method delivered what the abstract promises—reliable enforcement of a user-specified feature-space topology with limited impact on task performance—it would be a useful contribution, particularly for dimensionality reduction and representation learning. The paper has strengths: it is evaluated on three modalities, with a public implementation, repeated trials, and a clear baseline comparison. The computational overhead is modest and quantified. However, the central claim is weakened by two issues. First, the loss directly optimizes the measured statistic, so observing that H moves toward H_T is partly a tautology; the non-circular value lies in the downstream accuracy and reconstruction trade-offs, which are reported but not analyzed against a comparable control. Second, the empirical results fail to support the 'enforce' language for two of the three target topologies. The paper's own data in Figures 2 and 3 show mean H values that remain far outside the stated ranges for H_T = 0.01 and H_T = 0.5. A revision that re-frames the contribution as partial modification of clustering tendency, rather than topology enforcement, could make the claims sound.

major comments (4)
  1. [Section V, Figs. 2 and 3, Eq. (2)] The central claim of 'enforce[ing] a desired feature space topology' is not supported by the reported H values. For H_T = 0.01 (regularly spaced, H in [0.01,0.3]), the mean H after training ranges 0.52–0.73 across all experiments, still in the clustered or intermediate range. For H_T = 0.5, mean H ranges 0.56–0.80, not close to the target. Only H_T = 0.99 is approximated. The authors acknowledge in Section V that L_H 'could not force the features into a regularly-spaced topology,' but this is not reflected in the abstract or introduction. Please either provide formal evidence of target attainment (e.g., confidence intervals against the target intervals) or revise the claims throughout to 'partially modify' or 'shift' H, and adjust the title/abstract accordingly.
  2. [Section II-A, Eq. (2)] The Hopkins statistic is a scalar nearest-neighbor ratio that measures clustering tendency; it is not a general descriptor of feature-space topology. The paper equates H ranges with distinct 'topologies' (regularly spaced, randomly spaced, clustered), but many completely different point configurations yield the same H. Minimizing |H - H_T| therefore controls one scalar statistic, not the geometric topology in the sense used by the persistent-homology methods cited in Section II-B. The authors should either narrow the claims to 'clustering tendency' or provide evidence that H captures the intended geometric structure, e.g., qualitative scatter plots for all target values and all datasets, not just the single 2D example in Figure 4.
  3. [Section III, Eq. (2), Section V] Because L_H is defined as the absolute difference between the measured H and the target H_T, observing that H moves toward H_T is a direct optimization effect, not an independent validation of the method. The non-circular content is the downstream classification performance and reconstruction trade-off. To strengthen the claim that the method has practical 'benefit,' the paper should compare against a control that optimizes a different objective with comparable capacity, or at least report a systematic trade-off analysis (e.g., accuracy versus H for varying w_C and w_R). As written, the statement 'providing the benefit of modifying feature topology' is not quantified against a baseline that spends the same training budget on another regularizer.
  4. [Section III, Section IV] Several design choices are justified only by reference to 'preliminary experiments' without details. In particular, the selection of Chebyshev distance is load-bearing: the paper states that 'only Chebyshev distance was able to maintain the desired properties of H across all tested conditions,' but no supporting results are shown. Similarly, the choices m = 0.05n, w_C = 0.75, and w_R = 0.75 are presented without a sensitivity analysis. Please provide the preliminary experiment details, or at least a compact sensitivity table, so that readers can assess how robust the reported behavior is to these choices.
minor comments (5)
  1. [Abstract, Section VI] Inconsistent terminology: the abstract and introduction say 'enforce,' while the conclusion and Section V say 'modify.' Align the wording with the actual results.
  2. [Section IV-B, Fig. 2] The Mann-Whitney U tests are repeated for multiple H_T configurations and datasets without correction for multiple comparisons. Report whether the significance markers survive a multiple-testing correction, or state that they are uncorrected.
  3. [Section V, Figs. 2 and 3] The mean H values are reported with 95% confidence intervals that are often ±0.00 or ±0.01. Reporting standard deviations or effect sizes would give a clearer picture of variability, especially since the claim is about H moving toward H_T.
  4. [Section III] The sentence 'By calculating H using differentiable operations, H can be integrated into a loss function' is vague. Please specify how the min over nearest neighbors and the random sampling of Y are handled during backpropagation (e.g., straight-through or subgradient treatment).
  5. [Section II-A] Reference [23] is used to justify the H ranges for regularly spaced, randomly spaced, and clustered data. Since those ranges were established for simulated 2D data, state explicitly that the paper applies them to higher-dimensional feature spaces and note any caveats.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity found; the H-target evaluation is the explicit loss objective, not a hidden reduction.

full rationale

The paper's derivation chain is self-contained: it defines the Hopkins statistic H as a differentiable quantity and then explicitly sets the loss to LH = |H - HT| (Section III, Eq. 2). The reported movement of H toward HT is therefore the intended optimization objective, not a fitted parameter disguised as an independent prediction. The non-circular, load-bearing evidence is the effect on classification accuracy (Fig. 2), on autoencoder bottleneck classification (Fig. 3), and the qualitative visualizations (Fig. 4), none of which are implied by the loss definition alone. The paper also acknowledges in Section V that LH 'could not force the features into a regularly-spaced topology' with HT = 0.01, which demonstrates empirically that the H outcome is not a mathematical tautology; it depends on the multi-objective trade-off and optimization difficulty. The concern that H is only a scalar clustering-tendency statistic and may not capture geometric topology is a validity/correctness issue, not a circularity of the derivation. There are no load-bearing self-citations: the cited H-value ranges [23] and sampling heuristics [29,30] are external, and the paper does not invoke any uniqueness theorem from the authors' own prior work. Under the paper's own operational definition of topology in terms of H ranges, the loss-to-evaluation link is by design and is stated transparently. Thus no significant circularity is present.

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

The central claim rests on four hand-chosen design decisions (loss weights, distance metric, sample fraction, minibatch proxy) and one strong domain assumption that a scalar clustering-tendency index adequately describes feature-space topology. No new entities are postulated.

free parameters (4)
  • wC (classification loss weight) = 0.75
    Chosen from preliminary experiments for classification; balances LCE and LH.
  • wR (reconstruction loss weight) = 0.75
    Chosen from preliminary experiments for AE training; balances LMSE and LH.
  • Distance metric D = Chebyshev
    Selected in preliminary experiments because it was the only tested metric that preserved the expected H ranges across simulated conditions; this choice is tied to the method's validity.
  • Sample fraction m/n = 0.05
    Selected from preliminary experiments and Dubes and Jain recommendation; affects the H estimate's stability.
assumptions (4)
  • domain assumption H ranges correspond to distinct topologies: H in [0.01,0.3] means regularly spaced, H about 0.5 random, H in [0.7,0.99] clustered.
    Inherited from Banerjee and Dave [23] and used throughout the loss design and result interpretation (Section II-A, Figure 1). This is a scalar statistic, so many different layouts can share the same H.
  • domain assumption Minibatch H is a reliable proxy for the topology of the full feature set.
    The loss is computed on 1024-sample minibatches and m=0.05n subsamples; the paper does not validate that minibatch H matches full-dataset H (Section III, IV-B).
  • ad hoc to paper Gradient descent on LH with a piecewise-constant/min-based nearest-neighbor distance effectively optimizes H.
    The paper treats H as differentiable without specifying how the min operations are handled (ties, zero gradients); no subgradient analysis is provided (Section III).
  • ad hoc to paper The selected Chebyshev distance is appropriate for all tested feature spaces.
    Chebyshev was chosen after a search over five metrics based on simulated H behavior; this is a selection made by the authors, not an independent theoretical guarantee (Section III).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Feature Space Topology Control via Hopkins Loss." pith.science (2026). https://pith.science/paper/KGOFFLNI

@misc{pith2026250911154,
  author       = {Pith},
  title        = {Pith review of: Feature Space Topology Control via Hopkins Loss},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/KGOFFLNI}},
  note         = {Machine review of arXiv:2509.11154}
}
read the original abstract

Feature space topology refers to the organization of samples within the feature space. Modifying this topology can be beneficial in machine learning applications, including dimensionality reduction, generative modeling, transfer learning, and robustness to adversarial attacks. This paper introduces a novel loss function, Hopkins loss, which leverages the Hopkins statistic to enforce a desired feature space topology, which is in contrast to existing topology-related methods that aim to preserve input feature topology. We evaluate the effectiveness of Hopkins loss on speech, text, and image data in two scenarios: classification and dimensionality reduction using nonlinear bottleneck autoencoders. Our experiments show that integrating Hopkins loss into classification or dimensionality reduction has only a small impact on classification performance while providing the benefit of modifying feature topology.

Figures

Figures reproduced from arXiv: 2509.11154 by the authors.

Figure 1
Figure 1. An example 2D visualization of regularly-spaced (left, [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. The results of the classification experiments. The Hopkins loss target [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. The results of the AE experiments (left: bottleneck feature dimensionality [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Example of a 0.10 difference in H value when bottleneck feature dimensionality B = 2 for the RAVDESS dataset. The bottleneck features of a randomly initialized model (left, H = 0.89), a trained model with LH and HT = 0.5 (middle, H = 0.79), and a trained model with LH …

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

40 extracted references · 4 linked inside Pith

  1. [1]

    Rethinking Feature Distribution for Loss Functions in Image Classification,

    W. Wan, Y . Zhong, T. Li, and J. Chen, “Rethinking Feature Distribution for Loss Functions in Image Classification,” inProc. IEEE CVPR, 2018, pp. 9117–9126

    2018

  2. [2]

    Topology-Preserving Deep Image Segmentation,

    X. Hu, L. Fuxin, D. Samaras, and C. Chen, “Topology-Preserving Deep Image Segmentation,” inProc. NeurIPS, 2019, pp. 5657–5668

    2019

  3. [3]

    Dimensionality Reduction by Learning an Invariant Mapping,

    R. Hadsell, S. Chopra, and Y . LeCun, “Dimensionality Reduction by Learning an Invariant Mapping,” inProc. IEEE CVPR, 2006, pp. 1735– 1742

    2006

  4. [4]

    Improving Dimensionality Reduction Projections for Data Visualization,

    B. Rafieian, P. Hermosilla, and P.-P. V´azquez, “Improving Dimensionality Reduction Projections for Data Visualization,”Applied Sciences, vol. 13, no. 17, 2023

    2023

  5. [5]

    Wang,Dimensionality Reduction Task

    W. Wang,Dimensionality Reduction Task. Springer Nature Singapore, 2024, pp. 481–505

    2024

  6. [6]

    An Automated Data Mining Framework Using Autoencoders for Feature Extraction and Dimensionality Reduction,

    Y . Liang, X. Li, X. Huang, Z. Zhang, and Y . Yao, “An Automated Data Mining Framework Using Autoencoders for Feature Extraction and Dimensionality Reduction,”arXiv preprint arXiv: 2412.02211, 2024

    arXiv 2024

  7. [7]

    Generative Adversarial Nets,

    I. Goodfellow, J. Pouget-Abadie, M. Mirza, B. Xu, D. Warde-Farley, S. Ozair, A. Courville, and Y . Bengio, “Generative Adversarial Nets,” in Proc. NeurIPS, 2014, pp. 2672–2680

    2014

  8. [8]

    Controllable Generative Adversarial Network,

    M. Lee and J. Seok, “Controllable Generative Adversarial Network,” IEEE Access, vol. 7, pp. 28 158–28 169, 2019

    2019

Show all 40 references
  1. [9]

    Focal Frequency Loss for Image Reconstruction and Synthesis,

    L. Jiang, B. Dai, W. Wu, and C. C. Loy, “Focal Frequency Loss for Image Reconstruction and Synthesis,” inProc. IEEE ICCV, 2021, pp. 13 919–13 929

    2021

  2. [10]

    How Well Generative Adversarial Networks Learn Distri- butions,

    T. Liang, “How Well Generative Adversarial Networks Learn Distri- butions,”Journal of Machine Learning Research, vol. 22, no. 228, pp. 10 366–10 406, 2021

    2021

  3. [11]

    ProFET: Feature engineering captures high-level protein functions,

    D. Ofer and M. Linial, “ProFET: Feature engineering captures high-level protein functions,”Bioinformatics, vol. 31, no. 21, pp. 3429–3436, 2015

    2015

  4. [12]

    Taxonomy-aware feature engineering for microbiome classification,

    M. Oudah and A. Henschel, “Taxonomy-aware feature engineering for microbiome classification,”BMC Bioinformatics, vol. 19, no. 227, 2018

    2018

  5. [13]

    Transfer Learning across Feature-Rich Hetero- geneous Feature Spaces via Feature-Space Remapping (FSR),

    K. Feuz and D. Cook, “Transfer Learning across Feature-Rich Hetero- geneous Feature Spaces via Feature-Space Remapping (FSR),”ACM Transactions on Intelligent Systems and Technology, vol. 6, no. 1, pp. 1–27, 2015

    2015

  6. [14]

    Evaluation of Transfer Learning Algorithms using Different Base Learners,

    K. Weiss and T. Khoshgoftaar, “Evaluation of Transfer Learning Algorithms using Different Base Learners,” inProc. IEEE ICTAI, 2017, pp. 187–196

    2017

  7. [15]

    Un- supervised Domain Adaptation using Feature-Whitening and Consensus Loss,

    S. Roy, A. Siarohin, E. Sangineto, S. R. Bulo, N. Sebe, and E. Ricci, “Un- supervised Domain Adaptation using Feature-Whitening and Consensus Loss,” inProc. IEEE CVPR, 2019, pp. 9471–9480

    2019

  8. [16]

    Transfer Learning across Feature-Rich Heterogeneous Feature Spaces via Feature- Space Remapping (FSR),

    S. Khan, P. Yin, Y . Guo, M. Asim, and A. A. A. El-Latif, “Transfer Learning across Feature-Rich Heterogeneous Feature Spaces via Feature- Space Remapping (FSR),”Multimedia Tools and Applications, vol. 83, pp. 69 759–69 795, 2024

    2024

  9. [17]

    A Topology Layer for Machine Learning,

    R. Br ¨uel-Gabrielsson, B. Nelson, A. Dwaraknath, and P. Skraba, “A Topology Layer for Machine Learning,” inProc. AISTATS, 2020, pp. 1553–1563

    2020

  10. [18]

    Delving into Feature Space: Improving Adversarial Robustness by Feature Spectral Regular- ization,

    Z. Cheng, F. Zhu, X.-Y . Zhang, and C.-L. Liu, “Delving into Feature Space: Improving Adversarial Robustness by Feature Spectral Regular- ization,” inProc. ICPR, 2024, pp. 423–439

    2024

  11. [19]

    Enhancing the Resilience of Graph Neural Networks to Topological Perturbations in Sparse Graphs,

    S. He, J. Zhuang, D. Wang, L. Peng, and J. Song, “Enhancing the Resilience of Graph Neural Networks to Topological Perturbations in Sparse Graphs,”arXiv preprint arXiv: 2406.03097, 2024

    2024 arXiv

  12. [20]

    A New Method for determining the Type of Distribution of Plant Individuals,

    B. Hopkins and J. Skellam, “A New Method for determining the Type of Distribution of Plant Individuals,”Annals of Botany, vol. 18, no. 2, pp. 213–227, 1954

    1954

  13. [21]

    New index for clustering tendency and its application to chemical problems,

    R. Lawson and P. Jurs, “New index for clustering tendency and its application to chemical problems,”Journal of Chemical Information & Computer Sciences, vol. 30, no. 1, pp. 36–41, 1990

    1990

  14. [22]

    Measurement of Clustering Tendency,

    G. Cross and A. Jain, “Measurement of Clustering Tendency,”IFAC Proceedings Volumes, vol. 15, no. 1, pp. 315–320, 1982

    1982

  15. [23]

    Validating clusters using the Hopkins statistic,

    A. Banerjee and R. Dav´e, “Validating clusters using the Hopkins statistic,” inProc. IEEE FUZZ, 2004, pp. 149–153

    2004

  16. [24]

    Awareness of Line- of-Sight Propagation for Indoor Localization Using Hopkins Statistic,

    Z. Li, Z. Tian, M. Zhou, Z. Zhang, and Y . Jin, “Awareness of Line- of-Sight Propagation for Indoor Localization Using Hopkins Statistic,” IEEE Sensors Journal, vol. 18, no. 9, pp. 3864–3874, 2018

    2018

  17. [25]

    A novel density peaks clustering algorithm based on Hopkins statistic,

    R. Zhang, Z. Miao, Y . Tian, and H. Wang, “A novel density peaks clustering algorithm based on Hopkins statistic,”Expert Systems with Applications, vol. 201, p. 116892, 2022

    2022

  18. [26]

    A topological loss function for image Denoising on a new BVI-lowlight dataset,

    A. Malyugina, N. Anantrasirichai, and D. Bull, “A topological loss function for image Denoising on a new BVI-lowlight dataset,”Signal Processing, vol. 211, no. 109081, 2023

    2023

  19. [27]

    Manifold Learning by Mixture Models of V AEs for Inverse Problems,

    G. S. Alberti, J. Hertrich, M. Santacesaria, and S. Sciutto, “Manifold Learning by Mixture Models of V AEs for Inverse Problems,”Journal of Machine Learning Research, vol. 25, no. 202, pp. 1–35, 2024

    2024

  20. [28]

    Topology-preserving Adversarial Training for Alleviating Natural Accuracy Degradation,

    X. Mi, F. Tang, Y . Weng, D. Wang, J. Cao, S. Tang, P. Li, and Y . Liu, “Topology-preserving Adversarial Training for Alleviating Natural Accuracy Degradation,” inProc. BMVC, 2024

    2024

  21. [29]

    Modelling Spatial Patterns,

    B. D. Ripley, “Modelling Spatial Patterns,”Journal of the Royal Statistical Society. Series B (Methodological), vol. 39, no. 2, pp. 172–212, 1977

    1977

  22. [30]

    Clustering techniques: The user’s dilemma,

    R. Dubes and A. Jain, “Clustering techniques: The user’s dilemma,” Pattern Recognition, vol. 8, no. 4, pp. 247–260, 1976

    1976

  23. [31]

    The Ryerson Audio-Visual Database of Emotional Speech and Song (RA VDESS): A dynamic, multimodal set of facial and vocal expressions in North American English,

    S. Livingstone and F. Russo, “The Ryerson Audio-Visual Database of Emotional Speech and Song (RA VDESS): A dynamic, multimodal set of facial and vocal expressions in North American English,”PLoS ONE, vol. 13, no. 5, pp. 1–35, 2018

    2018

  24. [32]

    The Geneva Minimalistic Acoustic Parameter Set (GeMAPS) for V oice Research and Affective Computing,

    F. Eyben, K. Scherer, B. Schuller, J. Sundberg, E. Andr ´e, C. Busso, L. Devillers, J. Epps, P. Laukka, S. Narayanan, and K. Truong, “The Geneva Minimalistic Acoustic Parameter Set (GeMAPS) for V oice Research and Affective Computing,”IEEE Transactions on Affective Computing, ...

    2016

  25. [33]

    Recent developments in openSMILE, the Munich open-source multimedia feature extractor,

    F. Eyben, F. Weninger, F. Gross, and B. Schuller, “Recent developments in openSMILE, the Munich open-source multimedia feature extractor,” inProc. ACM, 2013, pp. 835–838

    2013

  26. [34]

    Learning Word Vectors for Sentiment Analysis,

    A. Maas, R. Daly, P. Pham, D. Huang, A. Ng, and C. Potts, “Learning Word Vectors for Sentiment Analysis,” inProc. ACL, 2011, pp. 142–150

    2011

  27. [35]

    BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding,

    J. Devlin, M. Chang, K. Lee, and K. Toutanova, “BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding,” inProc. NAACL-HLT, 2019, pp. 4171–4186

    2019

  28. [36]

    Transformers: State-of-the-Art Natural Language Processing,

    T. Wolf, L. Debut, V . Sanh, J. Chaumond, C. Delangue, A. Moi, P. Cistac, T. Rault, R. Louf, M. Funtowicz, J. Davison, S. Shleifer, P. von Platen, C. Ma, Y . Jernite, J. Plu, C. Xu, T. L. Scao, S. Gugger, M. Drame, Q. Lhoest, and A. M. Rush, “Transformers: State-of-the-Art Nat...

    2020

  29. [37]

    Fashion-MNIST: a Novel Image Dataset for Benchmarking Machine Learning Algorithms,

    H. Xiao, K. Rasul, and R. V ollgraf, “Fashion-MNIST: a Novel Image Dataset for Benchmarking Machine Learning Algorithms,”arXiv preprint arXiv: 1708.07747, 2017

    2017 arXiv

  30. [38]

    Gaussian Error Linear Units (GELUs),

    D. Hendrycks and K. Gimpel, “Gaussian Error Linear Units (GELUs),” arXiv preprint arXiv: 1606.08415, 2016

    2016 arXiv

  31. [39]

    Adam: A Method for Stochastic Optimization,

    D. Kingma and J. Ba, “Adam: A Method for Stochastic Optimization,” inProc. ICLR, 2015

    2015

  32. [40]

    On a Test of Whether one of Two Random Variables is Stochastically Larger than the Other,

    H. Mann and D. Whitney, “On a Test of Whether one of Two Random Variables is Stochastically Larger than the Other,”The Annals of Mathematical Statistics, vol. 18, no. 1, pp. 50–60, 1947

    1947

Pith tools

Reviewed August 4, 2026 · model on record in the stance chip above.