Pith. sign in

REVIEW 3 major objections 7 minor 40 references

Multiclass Classification without Labels via Posterior Simplex Geometry

T0 review · 3 major / 7 minor · reviewed 2026-07-31 · grok-4.5

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

Pith's one-line read Training only on which mixture a point came from recovers the hidden multiclass labels through a simplex in posterior space.

desk verdict Clean multiclass extension of CWoLa via posterior simplex geometry; theory is solid under stated assumptions, but every experiment engineers the two fragile assumptions that real cut-induced mixtures often break. read the letter →

arxiv 2607.24943 v1 pith:SAMRP2RU submitted 2026-07-27 cs.LG astro-ph.GAcs.AIstat.ML

Raphaël Bonnet-Guerrini , Johann Ioannou-Nikolaides , Troels Petersen , Vincenzo Piuri This is my paper · ORCID
classification cs.LGastro-ph.GAcs.AIstat.ML
keywords classificationwithoutlabelsmulticlassCWoLaposteriorsimplexmixturedemixingweaksupervisionvertexhuntinglabel-scarcelearning
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

When you cannot get instance-level class labels, you can often still gather several unlabeled batches that differ only in how much of each hidden class they contain. This paper shows that a standard classifier trained solely to predict which batch an example came from is enough to recover the latent classes. The Bayes-optimal batch classifier maps every point into a low-dimensional simplex whose corners are induced by the hidden classes through the unknown mixing weights. Once those corners are found—by fitting a simplex after training or by building a bottleneck into the network—you can read off both the class of each point and the class fractions inside each batch, without ever seeing labels or proportions. On MNIST, CIFAR-10, and Galaxy10 images the method closes much of the gap to fully supervised training, offering a practical route to multiclass discovery wherever labels are scarce but mixture diversity is available.

What carries the argument

Posterior simplex geometry: g⋆(x) = V α(x) embeds the data in a (K−1)-simplex in Δ^{M−1} whose vertices are the normalized columns of the unknown mixing matrix; barycentric coordinates on that simplex recover the latent class posterior, and the vertices themselves recover the mixture compositions.

What would settle it

Build mixtures that deliberately violate the anchor condition (every class always co-occurs with others, so the posterior cloud never reaches the vertices) while keeping shared class-conditionals and full-rank mixing: if post-hoc fitting and the bottleneck still recover accurate aligned class labels and mixing fractions, the central geometric claim fails.

Watch

Extended reading notes

Core claim

For a multiclass mixture model with shared class-conditionals and full-rank mixing, the Bayes-optimal mixture posterior g⋆(x) equals V α(x), where the columns of V are the class-induced vertices of a (K−1)-simplex inside mixture-posterior space and α(x) is the latent class posterior under the pooled effective prior. Under a separability (anchor) condition those vertices are exactly the extreme points of the support of g⋆(X), so both the latent posteriors and the hidden mixing matrix are identifiable up to permutation from mixture identity alone.

Load-bearing premise

Each hidden class must own some region of the data space that no other class reaches, so that points actually land on the corners of the simplex; without those pure anchors the corners cannot be identified from the cloud alone.

Editorial extensions

If this is right

  • Mixture identity alone is enough supervision to recover both instance-level multiclass labels and batch-level class proportions when the geometric conditions hold.
  • Adding more mixtures than classes (M > K) supplies redundant constraints that improve simplex recovery and raise aligned accuracy toward known-prior oracles.
  • Post-hoc vertex hunting on a trained mixture classifier and an architectural bottleneck that forces g = Vα are two interchangeable prior-free recovery routes.
  • The same posterior cloud can be used to discover the unknown number of latent classes K by comparing held-out reconstruction error and a shuffled-null gap statistic across candidate simplices.
  • Domains that naturally produce weakly enriched batches—different experimental cuts, instruments, clinics, or populations—can train multiclass models without instance labels or known priors.

Reading between the lines

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

  • The rectangular (M > K) regime is the practical sweet spot: extra mixtures act like free geometric regularizers, so practitioners should prefer many weak enrichments over few pure ones when labels are unavailable.
  • Because the method recovers Π as well as labels, mixture-composition estimation itself becomes a first-class scientific deliverable in fields where prevalence, not individual labels, is the quantity of interest.
  • If shared class-conditionals fail across batches, recovered “classes” may be batch artifacts; a cheap diagnostic would be to check whether simplex vertices stay stable when one mixture is held out.
  • Cardinality discovery from the posterior cloud suggests the pipeline can serve as an unsupervised class-discovery tool even when K is unknown a priori.
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 / 7 minor

Summary. The paper extends CWoLa from binary to multiclass learning from M unlabeled mixtures with K latent classes, where only mixture identity is observed. Under shared class-conditionals (A1), full-rank mixing (A2), separability/anchors (A3), and uniform mixture sampling (A4), the authors prove (Theorems 1–3, App. B) that the Bayes-optimal mixture posterior factorizes as g*(x)=Vα(x), with V the class-induced vertex matrix of a (K−1)-simplex in Δ^{M−1}, α the latent posterior under the pooled effective prior, and the vertices identifiable as extreme points of the posterior support, yielding Π up to permutation. Two prior-free recovery procedures are proposed — post-hoc simplex fitting on the trained M-way posterior cloud, and an architectural bottleneck g_θ=V̂α_θ — and evaluated on MNIST, Fashion-MNIST, CIFAR-10, and Galaxy10 DECaLS against prior-free baselines (OvR, KSBS-Demix) and known-prior oracles (Wei-CCM/RCM, oracle simplex). A K-discovery procedure via held-out reconstruction plus a gap statistic is given in App. C, and a reusable library is released. The proofs are correct and elementary; the empirical geometry prediction is visibly confirmed in Fig. 2. The principal weakness is that all experiments construct mixtures by sampling from labeled class pools, so A1 holds by construction and A3 is nearly free, while the advertised applications (cuts, sources, populations) plausibly violate A1 — a gap the paper acknowledges in Sec. 5 but never tests.

Significance. If the empirical program is strengthened, this is a useful and timely contribution: it gives the first clean multiclass generalization of CWoLa with exact identifiability conditions, a falsifiable geometric signature that is verified in practice, fair prior-free baselines, and a released library. The honesty about the relationship to separable NMF and the disclosure of concurrent work are commendable. The impact would be highest in scientific domains (HEP, astronomy) where mixture identity is observable but labels are not — precisely the regimes where the A1/A3 realism gap identified above matters most, so closing that gap empirically would substantially raise the paper's value.

major comments (3)
  1. [Sec. 4.1 and App. D (mixture construction) vs. Sec. 1 and Sec. 5] Every mixture in the paper is generated by drawing y~π_m and sampling uniformly from the labeled class-y training pool. This construction makes A1 (shared class-conditionals) hold tautologically and gives A3 (anchor regions) enormous mass for free, since each pool element is a genuine class-k exemplar. The abstract and Sec. 1, however, advertise mixtures 'selected by different cuts, sources, populations, or experimental conditions' — mechanisms that operate on observables and will generally shift p_k across mixtures, violating A1, in which case the shared factorization g*=Vα does not exist. Sec. 5 concedes this failure mode but no experiment probes it. The Galaxy10 experiment uses real images but the same synthetic latent-class sampling, so it validates the mixture mechanism only in its idealized form. Given that the paper's central practical claim is a tool for label-scarce domains wher
  2. [App. D, Eq. (14) (cyclic-template blending of Π)] When M≥K, each Dirichlet row of Π is blended with a fixed cyclic identity-like template (Eq. 14, λ=0.35 for MNIST/Fashion/CIFAR, λ=0.25 for Galaxy10). This biases Π toward near-diagonal, high-dominance structure, simultaneously protecting full column rank (A2) and creating strong per-mixture enrichment, so the headline experiments never probe generic mixing matrices. App. D states performance is stable when max_k π_mk > 0.4, but the blend itself is what keeps draws in that regime. Please report results with unblended Dirichlet draws (λ=0) at the main (K,M) settings — at minimum CIFAR-10 K=M=10 and K=10, M=20 — so the reader can separate the effect of the geometric method from the effect of the templated mixture design.
  3. [Table 1 and App. A (statistical reporting)] All results use 3 seeds, and the headline prior-free comparison has margins comparable to the noise: in Table 1, Simplex 0.584±0.051 vs. OvR 0.476±0.034 (std, not SEM), so the 10.8-point gap is roughly a 1.8σ effect under the paper's own normality assumption. App. A states error bars are 1σ standard deviations, while the Fig. 3 caption states shaded regions are one standard error — these are inconsistent and materially change interpretation of the curves. The claim 'improving substantially over OvR' (Sec. 4.3) is a headline empirical result; it should be backed by more seeds (or paired tests across shared mixture draws, which would tighten the comparison considerably) and a consistent, stated error-bar convention throughout.
minor comments (7)
  1. [App. D, Baselines paragraph] App. D 'Baselines' states 'OvR is the only prior-free baseline besides ours', but KSBS-Demix [Katz-Samuels et al., 2017] is prior-free and appears in Figs. 3 and 9. Please correct.
  2. [Sec. 4.1, paragraph on empirical analysis organization] The sentence 'accuracy is computed with respect to the mixture identity and so is informative while aligned accuracy corresponds to the accuracy with respect to the latent label' is garbled; please rewrite. Also, the Galaxy10 frozen-ResNet50 description is repeated twice in this section.
  3. [Fig. 2 caption] The posterior cloud and simplices are shown through a shared 2D PCA projection; PCA of simplex-constrained points can distort vertex/edge relationships. A brief note on how much variance the projection retains, or a corner-plot/distance-matrix supplement, would make the geometric claim visually verifiable.
  4. [App. C (K discovery)] Held-out reconstruction error overestimates K on Galaxy10 while the gap statistic recovers K_true; on MNIST overcomplete regimes the two signals err in opposite directions. Since K-discovery is advertised as a feature, please state an explicit decision rule combining the two signals and its failure modes, rather than presenting the signals separately.
  5. [Sec. 1, MultiCWoLa repository footnote, and App. A] The code link renders as 'hereMultiCWoLa repository' with no visible URL in the text. Please provide a working, persistent link (and ideally an anonymized or versioned archive) so the reproducibility claim can be checked.
  6. [Sec. 3.2, 'Scope and Failure Modes'] It would help to state explicitly that the M=K bottleneck reduces to the factorized network of VolMinNet [Li et al., 2021] only architecturally, and that the novelty for M>K is the rectangular regime; the current sentence in Sec. 3.2 is easy to misread as a claim of equivalence.
  7. [Theorem 2 / App. B.2 notation] In Eq. (11) the reweighting uses π^Y_k/c_k; since c_k = M·π̄_k, consider writing it in terms of π̄ for consistency with the 'effective prior' framing in Remark 1.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: posterior-simplex factorization follows from Bayes rule under stated mixture assumptions; experiments train only on mixture identity.

full rationale

The load-bearing claims (Theorems 1–3) are elementary consequences of the multiclass mixture model under A1–A4. Theorem 1 writes g⋆_m(x) = q_m(x)/∑_j q_j(x), substitutes q_m = ∑_k π_mk p_k, and rearranges to g⋆(x) = V α(x) with α_k ∝ c_k p_k; affine independence of the columns of V follows from rank(Π)=K. Theorem 2 identifies α with the pooled latent posterior under π̄_k = c_k/M and recovers it by the left inverse of V. Theorem 3 uses anchors (A3) so that g⋆ hits the vertices, which are then the extreme points of conv supp(g⋆_#μ_X). None of these steps defines the target in terms of itself, fits a free parameter to latent labels and renames it a prediction, or imports a uniqueness theorem from overlapping authors. Training uses only mixture identity m via cross-entropy; latent Y and Π are withheld from prior-free methods and appear solely in post-hoc Hungarian alignment and oracle baselines. Concurrent/related citations (binary CWoLa, separable NMF, VolMinNet, Wei-CCM) supply context, not a self-citation chain that forces the result. Experimental mixture construction that makes A1/A3 easy is an external-validity concern, not derivation circularity. The mathematical core is self-contained against its stated assumptions.

Assumptions & free parameters 4 free parameters · 7 assumptions · 2 invented entities

The central identifiability claim rests on a standard finite mixture model plus four explicit assumptions (shared conditionals, full-rank Π, anchors, uniform mixture sampling). No new physical entities are postulated. Free parameters are ordinary ML hyperparameters and synthetic mixture-construction knobs, not constants fitted to claim a universal law.

free parameters (4)
  • Dirichlet concentration α for synthetic Π = 0.8 / 0.5
    Controls mixture impurity in experiments (α=0.8 MNIST/CIFAR, 0.5 Galaxy10); chosen by hand and affects difficulty.
  • Cyclic template blend λ = 0.35 (vision) / 0.25 (Galaxy10)
    Stabilizes rank when M≥K by mixing Dirichlet rows with identity-like template; hand-chosen (0.35 / 0.25).
  • Target latent cardinality K in recovery = assumed known or selected
    Main experiments assume K known; discovery procedure treats K as selectable but still a discrete choice over [2,M].
  • Simplex fitter / bottleneck regularization choices
    Post-hoc fitter family, entropy/temperature/slack, warm-up—standard training knobs that affect empirical recovery.
assumptions (7)
  • domain assumption A1 Shared class-conditionals: p_k independent of mixture m
    Makes factorization g⋆=Vα well-posed; stated in Sec. 3. Violation is listed as primary failure mode.
  • domain assumption A2 rank(Π)=K (hence M≥K)
    Needed for (K−1)-dimensional simplex and left-invertibility of V.
  • domain assumption A3 Separability / anchors: each class has a pure support region
    Load-bearing for vertex identifiability (Theorem 3); standard in topic models but strong for natural images.
  • domain assumption A4 Uniform mixture sampling P(S=m)=1/M (or known frequencies)
    Normalization convenience; paper notes unequal sizes replace 1/M by empirical frequencies.
  • domain assumption Finite mixture model q_m=Σ_k π_mk p_k with row-stochastic Π
    Generative setup of Sec. 3; classical.
  • standard math Bayes-optimal M-way posterior and cross-entropy training approximate g⋆
    Standard statistical learning link from population risk to g⋆; finite-sample approximation is empirical.
  • standard math Affine independence of columns of V when rank(Π)=K
    Linear algebra step in Theorem 1 proof.
invented entities (2)
  • Effective prior π̄_k=c_k/M and barycentric weights α(x) independent evidence
    purpose: Identify the latent posterior of the pooled training population with the simplex coordinates of g⋆
    Not a new physical object; a derived reparameterization of the mixture model. Independent evidence is the algebraic identity in Theorem 1–2.
  • Architectural bottleneck g_θ(x)=V̂ α_θ(x) independent evidence
    purpose: Enforce simplex factorization at training time as an alternative to post-hoc vertex hunting
    Model architecture choice, analogous to VolMinNet-style factorizations; evaluated empirically.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Multiclass Classification without Labels via Posterior Simplex Geometry." pith.science (2026). https://pith.science/paper/SAMRP2RU

@misc{pith2026260724943,
  author       = {Pith},
  title        = {Pith review of: Multiclass Classification without Labels via Posterior Simplex Geometry},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/SAMRP2RU}},
  note         = {Machine review of arXiv:2607.24943}
}
abstract

In many classification problems, reliable instance-level labels are unavailable. However, it is often possible to construct weakly enriched unlabeled samples: datasets selected by different cuts, sources, populations, or experimental conditions that change latent class proportions without revealing them. Classification without Labels (CWoLa) shows that, in the binary case ($K=2$), a classifier trained to distinguish two impure mixtures with different class proportions can recover an optimal class discriminator without knowing the mixture proportions. We extend this principle to multiclass learning from several unlabeled mixtures ($K>2$), where the learner observes only mixture identity and neither latent class labels nor class-prior matrices. We prove that, for a multiclass mixture model, the Bayes-optimal mixture classifier $g^\star$ maps data points into a $(K-1)$-simplex embedded in mixture-posterior space. The $K$ vertices of this simplex are induced by the latent classes through the unknown mixing matrix. Leveraging this geometry, we propose prior-free procedures that train a standard classifier to distinguish mixture identities and then extract latent class structure using either post-hoc simplex fitting or a bottleneck architecture. Experiments on MNIST, CIFAR-10, and Galaxy10 DECaLS show that mixture identity alone can recover latent classes and their fractions in the mixture. By narrowing the gap between weakly supervised and fully supervised performance, we provide a mathematically grounded, scalable tool for multiclass discovery in label-scarce domains.

Figures

Figures reproduced from arXiv: 2607.24943 by the authors.

Figure 1
Figure 1. Overview of posterior simplex geometry via multiclass CWoLa. From unlabeled mixtures [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Posterior-simplex geometry in source-posterior space for CIFAR-10 with [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. Performance as a function of mixture complexity on CIFAR-10. Left: aligned accuracy [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: Robustness to mixture purity on Galaxy10 DECaLS. We construct cyclic-purity mixtures [PITH_FULL_IMAGE:figures/full_fig_p009_4.png]
Figure 5
Figure 5. Figure 5: Latent Class Cardinality Fitting on the Galaxy10 DECaLS dataset for [PITH_FULL_IMAGE:figures/full_fig_p017_5.png]
Figure 6
Figure 6. Figure 6: Aggregate robustness of candidate prior-free recovery configurations. Each row corre [PITH_FULL_IMAGE:figures/full_fig_p018_6.png]
Figure 7
Figure 7. Figure 7: Per-dataset decomposition of the prior-free recovery comparison. [PITH_FULL_IMAGE:figures/full_fig_p019_7.png]
Figure 8
Figure 8. Figure 8: Scaling with the number of mixtures on Galaxy10 DECaLS. We fix the number of [PITH_FULL_IMAGE:figures/full_fig_p019_8.png]
Figure 9
Figure 9. Figure 9: Robustness to mixture purity on CIFAR-10. We construct cyclic-purity mixtures with [PITH_FULL_IMAGE:figures/full_fig_p020_9.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

40 extracted references · 2 linked inside Pith

  1. [1]

    2015 , MONTH = May, DOI =

    Lecun, Yann and Bengio, Yoshua and Hinton, Geoffrey , URL =. 2015 , MONTH = May, DOI =

    2015

  2. [2]

    2005 , url=

    The mnist database of handwritten digits , author=. 2005 , url=

    2005

  3. [3]

    URL http://www.cs.toronto.edu/kriz/cifar

    Cifar-10 (canadian institute for advanced research) , author=. URL http://www.cs.toronto.edu/kriz/cifar. html , volume=

  4. [4]

    ArXiv , year=

    Fashion-MNIST: a Novel Image Dataset for Benchmarking Machine Learning Algorithms , author=. ArXiv , year=

  5. [5]

    Monthly Notices of the Royal Astronomical Society , year=

    Galaxy Zoo: morphologies derived from visual inspection of galaxies from the Sloan Digital Sky Survey , author=. Monthly Notices of the Royal Astronomical Society , year=

  6. [6]

    Galaxy Zoo DECaLS: Detailed visual morphology measurements from volunteers and deep learning for 314\,000 galaxies , journal =

    Walmsley, Mike and Lintott, Chris and G. Galaxy Zoo DECaLS: Detailed visual morphology measurements from volunteers and deep learning for 314\,000 galaxies , journal =. 2022 , month =. doi:10.1093/mnras/stab2093 , url =

    doi:10.1093/mnras/stab2093 2022

  7. [7]

    2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , year=

    Deep Residual Learning for Image Recognition , author=. 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , year=

    2016

  8. [8]

    ImageNet: A large-scale hierarchical image database , year=

    Deng, Jia and Dong, Wei and Socher, Richard and Li, Li-Jia and Kai Li and Li Fei-Fei , booktitle=. ImageNet: A large-scale hierarchical image database , year=

Show all 40 references
  1. [9]

    International Journal of Computer Vision , year=

    ImageNet Large Scale Visual Recognition Challenge , author=. International Journal of Computer Vision , year=

  2. [10]

    Proceedings of the VLDB Endowment

    Snorkel: Rapid Training Data Creation with Weak Supervision , author=. Proceedings of the VLDB Endowment. International Conference on Very Large Data Bases , year=

  3. [11]

    Darg, D. W. and others. Galaxy Zoo: the fraction of merging galaxies in the SDSS and their morphologies. Mon. Not. Roy. Astron. Soc. 2010. doi:10.1111/j.1365-2966.2009.15686.x. arXiv:0903.4937

    2010

  4. [12]

    Nature Communications , year=

    Annotation-efficient deep learning for automatic medical image segmentation , author=. Nature Communications , year=

  5. [13]

    ArXiv , year=

    A Simple Framework for Contrastive Learning of Visual Representations , author=. ArXiv , year=

  6. [14]

    2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year=

    Masked Autoencoders Are Scalable Vision Learners , author=. 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year=

    2022

  7. [15]

    Neural Information Processing Systems , year=

    Co-teaching: Robust training of deep neural networks with extremely noisy labels , author=. Neural Information Processing Systems , year=

  8. [16]

    and Nachman, Benjamin and Thaler, Jesse

    Metodiev, Eric M. and Nachman, Benjamin and Thaler, Jesse. Classification without labels: Learning from mixed samples in high energy physics. JHEP. 2017. doi:10.1007/JHEP10(2017)174. arXiv:1708.02949

    2017 arXiv

  9. [17]

    Physical Review D , year=

    Learning to classify from impure samples with high-dimensional data , author=. Physical Review D , year=

  10. [18]

    arXiv: High Energy Physics - Phenomenology , year=

    CWoLa Hunting: Extending the Bump Hunt with Machine Learning , author=. arXiv: High Energy Physics - Phenomenology , year=

  11. [19]

    In Defense of One-Vs-All Classification , author=. J. Mach. Learn. Res. , year=

  12. [20]

    Proceedings of the 26th Annual Conference on Learning Theory , pages =

    Scott, Clayton and Blanchard, Gilles and Handy, Gregory , title =. Proceedings of the 26th Annual Conference on Learning Theory , pages =. 2013 , editor =

    2013

  13. [21]

    Asian Conference on Machine Learning , year=

    Multi-class Classification from Multiple Unlabeled Datasets with Partial Risk Regularization , author=. Asian Conference on Machine Learning , year=

  14. [22]

    The Twelfth International Conference on Learning Representations , year=

    Consistent Multi-Class Classification from Multiple Unlabeled Datasets , author=. The Twelfth International Conference on Learning Representations , year=

  15. [23]

    ArXiv , year=

    Binary Classification from Multiple Unlabeled Datasets via Surrogate Set Classification , author=. ArXiv , year=

  16. [24]

    ArXiv , year=

    Learning from Label Proportions: A Mutual Contamination Framework , author=. ArXiv , year=

  17. [25]

    Decontamination of Mutual Contamination Models , author=. J. Mach. Learn. Res. , year=

  18. [26]

    2023 , eprint=

    Learning from Label Proportions with Instance-wise Consistency , author=. 2023 , eprint=

    2023

  19. [27]

    Journal of Machine Learning Research , volume =

    Decontamination of Mutual Contamination Models , author =. Journal of Machine Learning Research , volume =

  20. [28]

    Proceedings of the 17th International Conference on Neural Information Processing Systems , pages =

    Donoho, David and Stodden, Victoria , title =. Proceedings of the 17th International Conference on Neural Information Processing Systems , pages =. 2003 , publisher =

    2003

  21. [29]

    ArXiv , year=

    Computing a nonnegative matrix factorization -- provably , author=. ArXiv , year=

  22. [30]

    2012 IEEE 53rd Annual Symposium on Foundations of Computer Science , year=

    Learning Topic Models -- Going beyond SVD , author=. 2012 IEEE 53rd Annual Symposium on Foundations of Computer Science , year=

    2012

  23. [31]

    ArXiv , year=

    A Practical Algorithm for Topic Modeling with Provable Guarantees , author=. ArXiv , year=

  24. [32]

    ArXiv , year=

    Anchor-Free Correlated Topic Modeling: Identifiability and Algorithm , author=. ArXiv , year=

  25. [33]

    IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing , year=

    Hyperspectral Unmixing Overview: Geometrical, Statistical, and Sparse Regression-Based Approaches , author=. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing , year=

  26. [34]

    Journal of the American Statistical Association , year=

    Using SVD for Topic Modeling , author=. Journal of the American Statistical Association , year=

  27. [35]

    Archetypal Analysis , urldate =

    Adele Cutler and Leo Breiman , journal =. Archetypal Analysis , urldate =

  28. [36]

    How , title =

    Jesus Tordesillas and Jonathan P. How , title =. CoRR , volume =. 2020 , url =. 2010.10726 , timestamp =

    2020 arXiv

  29. [37]

    2016 , issue_date =

    Liu, Tongliang and Tao, Dacheng , title =. 2016 , issue_date =. doi:10.1109/TPAMI.2015.2456899 , journal =

    2016

  30. [38]

    International Conference on Machine Learning , year=

    Learning Noise Transition Matrix from Only Noisy Labels via Total Variation Regularization , author=. International Conference on Machine Learning , year=

  31. [39]

    ArXiv , year=

    Provably End-to-end Label-Noise Learning without Anchor Points , author=. ArXiv , year=

  32. [40]

    Simplex Demixing: Disentangling Multiple Light-Flavor Jets at Colliders , year =

Pith tools

Reviewed July 31, 2026 · model on record in the stance chip above.