Pith. sign in

REVIEW 3 major objections 5 minor 31 references

Improving Fine-Tuning with Latent Cluster Correction

T0 review · 3 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read Adding a latent-cluster correction loss improves fine-tuning accuracy on CIFAR-100 by about one point.

desk verdict A clearly written clustering-loss fine-tuning method, but the headline result evaporates once you notice that k=500 makes the loss empty on CIFAR-100; the remaining evidence is single-run and best-of-grid. read the letter →

arxiv 2501.11919 v1 pith:JGYBK2JH submitted 2025-01-21 cs.LG

Cédric Ho Thanh This is my paper · ORCID
classification cs.LG
keywords latentclustersfine-tuningclusteringlossLouvaincommunitydetectionCIFAR-100spacegeometryrepresentationlearningclassification
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 claims that a neural network's latent representations organize into semantic clusters, and that actively making those clusters match the true labels improves classification. It proposes latent cluster correction (LCC), a fine-tuning loss that detects clusters with a k-nearest-neighbor graph plus Louvain community detection, finds an optimal cluster-to-class assignment, and pulls misclustered samples toward the centroid of nearby correctly clustered samples from the same class. On CIFAR-100, adding this loss to ResNet-18 fine-tuning raises accuracy from 78.51% to 79.53% at the best setting, a gain of about one point, with consistent gains across most settings tested. The same loss on AlexNet gives negligible or slightly negative results, so the paper concludes the method is viable but architecture-dependent. The significance is that a loss shaping latent geometry, rather than a new architecture or data, can improve a standard fine-tuning pipeline.

What carries the argument

The machinery is the clustering loss of Eq. (8), a correction term that operates on latent representations: it builds a $k$-NN graph of the latent vectors, runs Louvain-Leiden community detection to obtain clusters, solves an optimal cluster-to-label assignment as a max-flow problem, and for each misclustered sample pulls it toward the centroid of its $k$ nearest correctly clustered same-class neighbors with the loss $\mathcal{L}_{\mathrm{clst}} = \frac{1}{\sqrt{d} N_{\mathrm{corr}}} \sum \|z_i - \bar{z}_i\|$. The Louvain modularity objective $Q$ in Eq. (2) is what lets the method discover an unspecified number of communities, avoiding any prior on the number of clusters; the max-flow assignment turns those communities into class-aligned correction targets. This pipeline is the object that carries the claim that reshaping latent geometry improves classification accuracy.

What would settle it

Re-run the best ResNet-18 configuration (head layer, $w=10^{-4}$, $k=500$) on CIFAR-100 across multiple seeds; if the +1.02 point gain over the 78.51% baseline does not reproduce, or if the gain persists when the 'correctly clustered' labels are replaced by random same-class centroids, the semantic-cluster premise would be falsified.

Watch

Extended reading notes

Core claim

Latent cluster correction (LCC) is a fine-tuning procedure built on the observation that correctly clustered samples are consistently more accurate than the overall set, and misclustered samples are worse. The loss $\mathcal{L}_{\mathrm{clst}}$ (Eq. 8) is the normalized mean over correctible misclustered samples of $\|z_i - \bar{z}_i\|$, where $\bar{z}_i$ is the centroid of the $k$ nearest neighbors of $z_i$ among the correctly clustered samples of the same true class. Clusters are found by Louvain-Leiden community detection on the $k$-NN graph of latent representations, and matched to true labels by an optimal one-to-many assignment computed with a max-flow formulation. The paper's central empirical claim is that adding this loss to standard cross-entropy fine-tuning on CIFAR-100 yields appreciable and consistent accuracy gains for ResNet-18, with the best configuration reaching 79.53% accuracy against a 78.51% baseline, while AlexNet shows no reliable gain. The author also reports an associated regularity: before and during training, accuracy on correctly clustered samples is higher than overall accuracy, which is higher than accuracy on misclustered samples, with the gap widening in deeper layers.

Load-bearing premise

The load-bearing premise is that the clusters detected by Louvain in the raw latent space align with true class boundaries closely enough that the centroid of neighboring correctly clustered samples is a trustworthy correction target; if the partition splits a class arbitrarily or merges classes, the loss pulls samples in the wrong direction.

Editorial extensions

If this is right

  • ResNet-18 fine-tuned on CIFAR-100 with LCC improves from 78.51% to 79.53% accuracy in the best configuration (head layer, $w=10^{-4}$, $k=500$), and every tested ResNet-18 setting beats the baseline.
  • Because the loss needs no preset number of clusters, LCC can be applied to latent spaces where one class splits into several clusters, unlike methods that force one cluster per class.
  • The rank ordering $\mathrm{acc}(\mathrm{MC}) < \mathrm{acc}(\mathrm{all}) < \mathrm{acc}(\mathrm{CC})$ gives a cheap diagnostic: if a dataset's latent space does not show it, LCC is unlikely to help.
  • AlexNet's mostly negative results imply LCC should be selected per architecture, not applied as a general default in fine-tuning.
  • The method's $O(kN^2)$ worst-case Louvain step and full-dataset memory requirement confine current use to small datasets such as CIFAR-100.

Reading between the lines

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

  • Beyond the paper: if cluster alignment is the causal mechanism behind the gain, then LCC should transfer to other label-structured tasks, such as few-shot classification or out-of-distribution detection, where latent clusters are known to be poorly aligned.
  • Beyond the paper: the architecture dependence suggests a testable criterion for when to apply LCC — measure the CC/MC accuracy gap on the chosen layer and only correct when the gap is present; ResNet-18's gains and AlexNet's null results should track this gap.
  • Beyond the paper: a natural extension is to make the correction target class-dependent and cluster-aware, e.g. using the orthogonal projection proposed in the discussion (Section V.C), which should speed separation of misclustered samples from the wrong object manifold.
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

3 major / 5 minor

Summary. The paper introduces Latent Cluster Correction (LCC), a fine-tuning method that detects latent-space clusters via k-NN-Louvain community detection, matches clusters to true labels through a max-flow formulation, and adds a clustering loss that pulls misclustered samples toward centroids of correctly clustered same-class samples. Experiments on CIFAR-100 with three ImageNet-pretrained architectures (TinyNet, ResNet-18, AlexNet) report accuracy gains for ResNet-18 but mixed or negative results for AlexNet. The central claim is that adding L_clst (Eq. 8) to cross-entropy fine-tuning improves classification accuracy.

Significance. If the central claim were robustly supported, LCC would be a potentially interesting addition to the fine-tuning toolkit, particularly for its use of community detection to handle non-spherical clusters and its explicit handling of one-to-many cluster-to-class matching. The max-flow assignment in Section III.C is clearly presented and comes with a proof, and the paper is honest about the method's computational cost and preliminary nature. However, as it stands, the experimental evidence is not convincing: the best reported result is produced by a configuration for which the clustering loss is identically zero, the remaining gains come from single runs with no error bars and are selected from a 12-configuration grid, and the architecture-dependence claim is contradicted by the mostly negative AlexNet results. These issues preclude acceptance without substantial revision.

major comments (3)
  1. [Section III.D, Eq. (8); Table III, Head row, k=500] The definition of a correctible sample requires k CC samples in the same true class. On CIFAR-100, each class has exactly 500 training samples. Since a misclustered sample is not CC, the maximum number of CC samples available in its class is 499, which is less than k=500. Therefore no sample is correctible, N_corr=0, and L_clst is identically zero (or undefined if one does not adopt the convention that an empty sum is zero). The ResNet-18 accuracy of 79.53% reported for Head, w=1e-4, k=500 cannot be attributed to the clustering loss; it is either a run-to-run artifact or an undocumented violation of Eq. (8). This invalidates the headline +1.02% gain and weakens the claim of consistent improvements, since the valid k=5 and k=50 rows are all single runs.
  2. [Section IV.B, Tables III and IV] Each reported accuracy is a single run with no variance, confidence interval, or number of seeds. The best configuration is then selected from a grid of 12 configurations (2 layers × 2 values of w × 3 values of k). Under multiple comparisons, the maximum of 12 noisy estimates is biased upward; the reported gains, including the remaining +0.73% and +0.95% for k=5/50, may be within run-to-run noise. The paper should report mean and standard deviation over several seeds, and compare methods with paired tests or confidence intervals on the differences.
  3. [Abstract, Section IV.B Table IV, Section VII] The abstract and conclusion claim that LCC boosts performance, but the AlexNet results in Table IV are predominantly negative: only 2 of 12 configurations show a positive gain, with a maximum of +0.23%, while many are negative, including -0.58%. The text acknowledges 'lukewarm' results but does not reconcile this with the general claim of viability. The conclusion's statement that 'on average, LCC can noticeable classification accuracy improvements' is not supported by the AlexNet data, where the average is negative. The claim should be narrowed to specific architectures or the experimental evidence should be expanded to show when LCC helps or hurts.
minor comments (5)
  1. [Section III.C, Eq. (5)] Eq. (5) uses 'y_n' in the set-builder notation; it should be 'y_i' to be consistent with the definition of the dataset (Z, y).
  2. [Section III.D, Eq. (8)] The loss definition does not specify the convention when N_corr=0. Given the k=500 issue, the paper must explicitly state that L_clst is defined as 0 when no correctible samples exist, and then ensure the reported configuration is actually consistent with that convention.
  3. [Section II, Table I] The comparison table lists properties such as 'Handles non-linearly sep. clsts.' for spectral clustering as ×, but spectral clustering can handle non-linearly separable clusters via kernels; this at least needs a qualifier or reference.
  4. [Section VI.A, Eq. (12)] The conductance update equations contain several notation errors: 'v_j = 2 ∑_i A B^{(r)}_{i,u}' should presumably index over j, not u, and the definition of u as a vector is tangled with the cut calculation. The pseudo-code is very hard to follow and should be rewritten with clear indexing.
  5. [Throughout] Numerous typos and misspellings: 'hundreads', 'Nonetheness', 'eventhough', 'discrepency', 'missclustered' in figure captions, 'prepresentations', 'wether', 'proceduce', 'begining'. A careful proofreading pass is needed.

Circularity Check

1 steps flagged · score 6.0 of 10

The paper's best ResNet-18 gain comes from a k=500 run where, on CIFAR-100, the clustering loss is identically zero by Eq. (7), so the headline result is not attributable to LCC.

  1. other [Section III.D Eqs. (7)-(8); Section IV.B Table III row 'Head 10^-4 500']
    "'We say that a MC sample z_i is correctible if there exists k CC samples z_{j1}, …, z_{jk} in y_i. In this case, the target of z_i is the centroid z̄_i = (z_{j1} + ⋯ + z_{jk})/k' and 'ℒ_clst = 1/(√d N_corr) ∑_{z_i corr.} ‖z_i − z̄_i‖'. Table III lists 'Head 10^-4 500 79.53% +1.02%'."

    On CIFAR-100, each of the 100 classes has exactly 500 training samples. A misclustered (MC) sample is by definition not correctly clustered (CC), so even if every other sample in its true class is CC, only 499 CC samples exist in that class. Eq. (7) requires k=500 CC samples for correctibility, which is impossible. Therefore N_corr=0 and the mean in Eq. (8) is empty, so L_clst provides no gradient. The k=500 experiments are thus standard fine-tuning without the proposed loss, yet Table III reports the largest gain (+1.02%) for that configuration and the text claims 'Applying LCC on ResNet-18 (Table III) shows appreciable and consistent accuracy gains.' The headline empirical 'prediction' reduces by the paper's own equations to a run where the clustering loss has no effect.

full rationale

The derivation of LCC is otherwise self-contained: the loss is defined directly from ground-truth labels, optimal cluster-label matching, and centroids of correctly clustered samples; there are no load-bearing self-citations or imported uniqueness theorems. The only concrete circular reduction is in the evaluation of the headline result. The paper's Eq. (7) makes a sample correctible only if k CC samples exist in its true class. On CIFAR-100, each class contains exactly 500 training samples; a MC sample is by definition not CC, so at most 499 CC samples can exist in its class. Hence for k=500 no sample is correctible, N_corr=0, and the Eq. (8) loss is an empty sum (or zero if guarded). The Table III rows with k=500 therefore run exactly the baseline fine-tuning procedure without any clustering-loss gradient, yet one of them (Head, w=10^-4) is reported as the best result, 79.53% (+1.02%), and the text concludes 'Applying LCC on ResNet-18 (Table III) shows appreciable and consistent accuracy gains.' That is a claimed prediction that reduces by construction to no LCC effect. The remaining k=5 and k=50 rows do exercise the loss, so the method is not vacuous overall, but they are single runs without error bars and the paper's 'consistent gains' statement selects the best of twelve configurations. Such selection is an evaluation weakness rather than a circular derivation; the zero-loss best row is the load-bearing circular step and warrants score 6.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The method rests on four main premises: semantic clusters exist and matter, Louvain on a k-NN graph finds them, pulling misclustered samples to CC centroids helps, and the loss gradient treats targets as constants. The first is a literature assumption, the second and third are assumed without quantitative validation, and the fourth is an unstated technical choice.

free parameters (3)
  • k (neighbors) = 5, 50, 500 (grid searched)
    Number of neighbors for the k-NN graph and correction targets; chosen per model and layer to maximize reported test accuracy.
  • w (clustering loss weight) = 1e-2, 1e-4 (grid searched)
    Weight of the clustering loss relative to cross-entropy; selected from the same grid as k.
  • Selected latent layer = classifier head or second-to-last trainable layer
    Which layer's representations are corrected; the paper reports results for both and picks the better one in the tables.
assumptions (4)
  • domain assumption Semantic clusters exist in latent spaces and their salience correlates with final accuracy.
    Stated in the abstract and Section I.A; the paper does not prove this correlation, it cites empirical observation.
  • ad hoc to paper Louvain community detection on a k-NN graph of latent representations yields meaningful, semantically relevant clusters.
    Section III.B relies on this for all downstream correction; the paper gives only a qualitative UMAP figure as support.
  • ad hoc to paper Pulling misclustered samples toward centroids of correctly clustered same-class samples reduces classification error.
    This is the design premise of the loss in Eq. (8); AlexNet's negative results in Table IV show it is not universally valid.
  • standard math The correction targets are treated as constants when differentiating the clustering loss.
    Section III.D defines the loss but does not describe backpropagation through the cluster assignment, implying detached targets.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Improving Fine-Tuning with Latent Cluster Correction." pith.science (2026). https://pith.science/paper/JGYBK2JH

@misc{pith2026250111919,
  author       = {Pith},
  title        = {Pith review of: Improving Fine-Tuning with Latent Cluster Correction},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/JGYBK2JH}},
  note         = {Machine review of arXiv:2501.11919}
}
read the original abstract

The existence of salient semantic clusters in the latent spaces of a neural network during training strongly correlates its final accuracy on classification tasks. This paper proposes a novel fine-tuning method that boosts performance by optimising the formation of these latent clusters, using the Louvain community detection algorithm and a specifically designed clustering loss function. We present preliminary results that demonstrate the viability of this process on classical neural network architectures during fine-tuning on the CIFAR-100 dataset.

Figures

Figures reproduced from arXiv: 2501.11919 by the authors.

Figure 1
Figure 1. LRs of 5000 samples from the CIFAR-10 dataset passed through a [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. On the left: LRs of 5000 samples from the Fashion MNIST [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Accuracy of correctly clustered and missclustered samples during the [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Accuracy of correctly clustered and missclustered samples during the [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

31 extracted references · 15 canonical work pages

  1. [1]

    Deep Residual Learning for Image Recognition,

    K. He, X. Zhang, S. Ren, and J. Sun, “Deep Residual Learning for Image Recognition,” in 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , Las Vegas, NV, USA: IEEE, Jun. 2016, pp. 770–778. doi: 10.1109/CVPR.2016.90

    doi:10.1109/cvpr.2016.90 2016

  2. [2]

    Learning Multiple Layers of Features from Tiny Images,

    A. Krizhevsky, “Learning Multiple Layers of Features from Tiny Images,” 2009

    2009

  3. [3]

    UMAP: Uniform Mani- fold Approximation and Projection for Dimension Reduction,

    L. McInnes, J. Healy, and J. Melville, “UMAP: Uniform Mani- fold Approximation and Projection for Dimension Reduction,” no. arXiv:1802.03426. arXiv, Sep. 2020

    arXiv 2020

  4. [4]

    Quantifying the Separability of Data Classes in Neural Networks,

    A. Schilling, A. Maier, R. Gerum, C. Metzner, and P. Krauss, “Quantifying the Separability of Data Classes in Neural Networks,” Neural Networks , vol. 139, pp. 278–293, Jul. 2021, doi: 10.1016/ j.neunet.2021.03.035

    work page 2021

  5. [5]

    Separability and Geometry of Object Manifolds in Deep Neural Networks,

    U. Cohen, S. Chung, D. D. Lee, and H. Sompolinsky, “Separability and Geometry of Object Manifolds in Deep Neural Networks,” Nature Communications, vol. 11, no. 1, p. 746, Feb. 2020, doi: 10.1038/ s41467-020-14578-5

    work page 2020

  6. [6]

    Introducing Graph Smoothness Loss for Training Deep Learning Architectures,

    M. Bontonou, C. Lassance, G. B. Hacene, V. Gripon, J. Tang, and A. Ortega, “Introducing Graph Smoothness Loss for Training Deep Learning Architectures,” in 2019 IEEE Data Science Workshop (DSW) , Minneapolis, MN, USA: IEEE, Jun. 2019, pp. 160–164. doi: 10.1109/ DSW.2019.8755559

    arXiv 2019

  7. [7]

    Representing Deep Neural Networks Latent Space Geometries with Graphs,

    C. Lassance, V. Gripon, and A. Ortega, “Representing Deep Neural Networks Latent Space Geometries with Graphs,” Algorithms, vol. 14, no. 2, p. 39, Jan. 2021, doi: 10.3390/a14020039

    work page doi:10.3390/a14020039 2021

  8. [8]

    Distilling the Knowledge in a Neural Network,

    G. Hinton, O. Vinyals, and J. Dean, “Distilling the Knowledge in a Neural Network,” no. arXiv:1503.02531. arXiv, Mar. 2015

    arXiv 2015

Show all 31 references
  1. [9]

    FitNets: Hints for Thin Deep Nets,

    A. Romero, N. Ballas, S. E. Kahou, A. Chassang, C. Gatta, and Y. Bengio, “FitNets: Hints for Thin Deep Nets,” no. arXiv:1412.6550. arXiv, Mar. 2015

    2015 arXiv

  2. [10]

    Robust Feature Space Separation for Deep Convolutional Neural Network Training,

    A. Sekmen, M. Parlaktuna, A. Abdul-Malek, E. Erdemir, and A. B. Koku, “Robust Feature Space Separation for Deep Convolutional Neural Network Training,” Discover Artificial Intelligence, vol. 1, no. 1, p. 12, Dec. 2021, doi: 10.1007/s44163-021-00013-1

    2021 doi

  3. [11]

    Grassmannian Learning: Embedding Geometry Awareness in Shallow and Deep Learning,

    J. Zhang, G. Zhu, R. W. Heath Jr., and K. Huang, “Grassmannian Learning: Embedding Geometry Awareness in Shallow and Deep Learning,” no. arXiv:1808.02229. arXiv, Aug. 2018

    2018 arXiv

  4. [12]

    Learning Embedding Space for Clustering From Deep Representations,

    P. Dahal, “Learning Embedding Space for Clustering From Deep Representations,” in 2018 IEEE International Conference on Big Data (Big Data), Seattle, WA, USA: IEEE, Dec. 2018, pp. 3747–3755. doi: 10.1109/BigData.2018.8622629

    2018

  5. [13]

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

    H. Xiao, K. Rasul, and R. Vollgraf, “Fashion-MNIST: A Novel Image Dataset for Benchmarking Machine Learning Algorithms,” no. arXiv:1708.07747. arXiv, Sep. 2017

    2017 arXiv

  6. [14]

    ImageNet Classification with Deep Convolutional Neural Networks,

    A. Krizhevsky, I. Sutskever, and G. E. Hinton, “ImageNet Classification with Deep Convolutional Neural Networks,” in Advances in Neural Information Processing Systems , F. Pereira, C. Burges, L. Bottou, and K. Weinberger, Eds., Curran Associates, Inc., 2012

    2012

  7. [15]

    Subspace Clustering for High Dimensional Data: A Review,

    L. Parsons, E. Haque, and H. Liu, “Subspace Clustering for High Dimensional Data: A Review,” ACM SIGKDD Explorations Newsletter, vol. 6, no. 1, pp. 90–105, Jun. 2004, doi: 10.1145/1007730.1007731

    2004

  8. [16]

    C. C. Aggarwal and C. K. Reddy, Eds., Data Clustering: Algorithms and Applications. in Chapman & Hall/CRC Data Mining and Knowledge Discovery Series. Boca Raton: Chapman and Hall/CRC, 2014

    2014

  9. [17]

    A Survey of Clustering With Deep Learning: From the Perspective of Network Architecture,

    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 , vol. 6, pp. 39501–39514, 2018, doi: 10.1109/ACCESS.2018.2855437

    2018

  10. [18]

    BIRCH: An Efficient Data Clustering Method for Very Large Databases,

    T. Zhang, R. Ramakrishnan, and M. Livny, “BIRCH: An Efficient Data Clustering Method for Very Large Databases,” ACM SIGMOD Record, vol. 25, no. 2, pp. 103–114, Jun. 1996, doi: 10.1145/235968.233324

    1996

  11. [19]

    Automatic Subspace Clustering of High Dimensional Data for Data Mining Appli- cations,

    R. Agrawal, J. Gehrke, D. Gunopulos, and P. Raghavan, “Automatic Subspace Clustering of High Dimensional Data for Data Mining Appli- cations,” in Proceedings of the 1998 ACM SIGMOD International Conference on Management of Data , Seattle Washington USA: ACM, Jun. 1998, pp. 94–...

    1998

  12. [20]

    OP- TICS: Ordering Points to Identify the Clustering Structure,

    M. Ankerst, M. M. Breunig, H.-P. Kriegel, and J. Sander, “OP- TICS: Ordering Points to Identify the Clustering Structure,” ACM SIGMOD Record , vol. 28, no. 2, pp. 49–60, Jun. 1999, doi: 10.1145/304181.304187

    1999

  13. [21]

    On Spectral Clustering: Analysis and an Algorithm,

    A. Y. Ng, M. I. Jordan, and Y. Weiss, “On Spectral Clustering: Analysis and an Algorithm,” in Proceedings of the 14th International Conference on Neural Information Processing Systems: Natural and Synthetic , in NIPS'01. Cambridge, MA, USA: MIT Press, 2001, pp. 849–856

    2001

  14. [22]

    Clustering by Passing Messages Between Data Points,

    B. J. Frey and D. Dueck, “Clustering by Passing Messages Between Data Points,” Science, vol. 315, no. 5814, pp. 972–976, Feb. 2007, doi: 10.1126/science.1136800

    2007 doi

  15. [23]

    Density- Based Clustering Based on Hierarchical Density Estimates,

    R. J. G. B. Campello, D. Moulavi, and J. Sander, “Density- Based Clustering Based on Hierarchical Density Estimates,” Advances in Knowledge Discovery and Data Mining , vol. 7819. Springer Berlin Heidelberg, Berlin, Heidelberg, pp. 160–172, 2013. doi: 10.1007/978-3-642-37456-2_14

    2013 doi

  16. [24]

    Community Detection Algorithms: A Comparative Analysis,

    A. Lancichinetti and S. Fortunato, “Community Detection Algorithms: A Comparative Analysis,” Physical Review E , vol. 80, no. 5, p. 56117, Nov. 2009, doi: 10.1103/PhysRevE.80.056117

    2009 doi

  17. [25]

    Fast Unfolding of Communities in Large Networks,

    V. D. Blondel, J.-L. Guillaume, R. Lambiotte, and E. Lefebvre, “Fast Unfolding of Communities in Large Networks,” Journal of Statistical Mechanics: Theory and Experiment , vol. 2008, no. 10, p. P10008, Oct. 2008, doi: 10.1088/1742-5468/2008/10/P10008

    2008 doi

  18. [26]

    From Louvain to Leiden: Guaranteeing Well-Connected Communities,

    V. A. Traag, L. Waltman, and N. J. Van Eck, “From Louvain to Leiden: Guaranteeing Well-Connected Communities,” Scientific Reports, vol. 9, no. 1, p. 5233, Mar. 2019, doi: 10.1038/s41598-019-41695-z

    2019 doi

  19. [27]

    Model Rubik's Cube: Twisting Resolution, Depth and Width for TinyNets,

    K. Han, Y. Wang, Q. Zhang, W. Zhang, C. Xu, and T. Zhang, “Model Rubik's Cube: Twisting Resolution, Depth and Width for TinyNets,” no. arXiv:2010.14819. arXiv, Dec. 2020

    2010 arXiv

  20. [28]

    Larochelle and Neural Information Processing Systems Foundation, Eds., 34th Conference on Neural Information Processing Systems (NeurIPS 2020): Online, 6-12 December 2020 , no

    H. Larochelle and Neural Information Processing Systems Foundation, Eds., 34th Conference on Neural Information Processing Systems (NeurIPS 2020): Online, 6-12 December 2020 , no. 33. in Advances in Neural Information Processing Systems. Red Hook, NY: Curran Associates, Inc, 2021

    2020

  21. [29]

    ImageNet: A Large-Scale Hierarchical Image Database,

    J. Deng, W. Dong, R. Socher, L.-J. Li, Kai Li, and Li Fei-Fei, “ImageNet: A Large-Scale Hierarchical Image Database,” in 2009 IEEE Conference on Computer Vision and Pattern Recognition , Miami, FL: IEEE, Jun. 2009, pp. 248–255. doi: 10.1109/CVPR.2009.5206848

    2009

  22. [30]

    Parallel Peer Pressure Clustering Algorithm Based on Linear Algebra Computation,

    J. Chen and P. Zou, “Parallel Peer Pressure Clustering Algorithm Based on Linear Algebra Computation,” Advanced Parallel Processing Technologies, vol. 10561. Springer International Publishing, Cham, pp. 105–116, 2017. doi: 10.1007/978-3-319-67952-5_10

    2017 doi

  23. [31]

    Implementing a Parallel Graph Clustering Algorithm with Sparse Matrix Computation,

    J. Chen and P. Zou, “Implementing a Parallel Graph Clustering Algorithm with Sparse Matrix Computation,” in 2018 IEEE Interna- tional Parallel and Distributed Processing Symposium Workshops (IPDPSW), Vancouver, BC: IEEE, May 2018, pp. 487–496. doi: 10.1109/IPDPSW.2018.00085

    2018

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.