Pith. sign in

REVIEW 4 major objections 7 minor 39 references

Fisher Information based Stochastic Gradient Ascent for Online Learning of Dirichlet Process Mixture and Theory

T0 review · 4 major / 7 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read A Fisher-information step size makes stochastic Bayesian clustering as accurate as closed-form inference.

desk verdict A useful adaptive-gradient heuristic for DPM variational inference, but the Fisher-information justification does not hold up; deserves revision, not desk rejection. read the letter →

arxiv 2412.08951 v3 pith:7AGXC4UQ submitted 2024-12-12 cs.LG stat.ML

Kart-Leong Lim , Xudong Jiang This is my paper · ORCID
classification cs.LGstat.ML
keywords DirichletprocessmixturestochasticgradientascentFisherinformationadaptivestepsizevariationalinferenceBayesiannonparametricsimageclusteringmodelselection
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 tries to establish that a Dirichlet process mixture—a Bayesian clustering model that also estimates the number of clusters—can be learned with stochastic gradient ascent rather than the usual closed-form coordinate ascent, without paying a performance penalty. The proposed update scales the gradient of the variational log-posterior by an empirical Fisher information term, which the authors argue is the correct direction for steepest ascent along the curvature of the posterior and gives an automatic adaptive stepsize. In experiments across six image datasets, the Fisher-scaled learner matches or beats the closed-form maximization–maximization baseline in normalized mutual information and accuracy, estimates a model closer to the ground-truth number of classes, and runs faster; on SUN397, with about 108,000 images and 397 classes, it is about 30% faster than the baseline. If correct, this removes the closed-form bottleneck that has limited scalable variational inference for Bayesian nonparametrics.

What carries the argument

The carrying object is an empirical Fisher-information scaling of the stochastic gradient. For each variational parameter $\theta$ (cluster mean $\mu_k$ or stick-breaking weight $v_k$), the update replaces the plain gradient step by one divided by $F'_\theta = \frac{1}{M}\sum_{n=1}^M (\nabla_\theta \ln q(\theta))^2$, the mean square of the minibatch gradient; with the diagonal-covariance assumption the Fisher term is a scalar per dimension, so the cost is just a per-coordinate division. This scaling does the work of adapting the stepsize to the local curvature: the derivation from Eq. (12)–(13) treats $F_\theta$ as the metric in a KL-divergence constraint, which is what makes the scaled gradient an approximate natural-gradient step rather than an arbitrary damping factor.

What would settle it

Run SGA+F on a synthetic Dirichlet process Gaussian mixture with a known posterior, record the effective stepsize $F_\theta^{-1}\eta$ for each coordinate, and compare it with the inverse diagonal of the true posterior covariance. If the ratio between the two varies strongly across coordinates, or diverges when the likelihood is flat, the empirical-Fisher metric is not the natural-gradient metric and the adaptive-step claim is refuted.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central claim is that Fisher information can be injected into stochastic gradient ascent for variational inference in Dirichlet process mixtures, turning a slow constant-stepsize learner into a fast adaptive one. Concretely, the variational expectations for the cluster means and stick-breaking weights are updated by $E[\theta] = E[\theta]' + F_\theta^{-1} \eta \nabla_\theta \ln q(\theta)$, where $F_\theta = E[(\nabla_\theta \ln q(\theta))^2]$ is the Fisher information of the variational log-posterior, computed empirically over each minibatch with negligible overhead since the gradient is already available. The authors derive this scaling from the KL-divergence constraint that makes $F_\theta$ the Riemannian metric for steepest ascent, and they position it as the first use of Fisher information in stochastic-gradient-ascent variational inference for a Dirichlet process mixture. Their experiments compare against the closed-form maximization–maximization learner and report on-par or better NMI and accuracy on all six datasets, better model selection on several, and lower CPU time, including on Caltech256 and SUN397 with VGG16 features.

Load-bearing premise

The load-bearing premise is that the empirical second moment of the gradient, $F_\theta = E[(\nabla_\theta \ln q(\theta))^2]$, is the right curvature metric for choosing the stepsize; if this scalar scaling does not capture the true curvature of the variational posterior, the adaptive-stepsize claim collapses and the method becomes a heuristic preconditioned gradient.

Editorial extensions

If this is right

  • The closed-form requirement for scalable variational inference in Dirichlet process mixtures is removed: any model where $\nabla_\theta \ln q(\theta)$ can be computed, even without an analytic expectation, becomes learnable by SGA+F.
  • Adaptive stepsize via Fisher information should converge faster and with less fluctuation than constant-stepsize stochastic gradient ascent on the same objective, as the convergence plots in the paper indicate.
  • The method inherits the scalability of minibatch training, so a Dirichlet process mixture can be applied to datasets at the scale of SUN397 (108K images, 397 classes) that batch closed-form learners cannot fully see.
  • For the six datasets tested, SGA+F should deliver NMI and accuracy at least as good as the closed-form MM baseline while using less CPU time.
  • High-dimensional deep features (4096-dimensional VGG16) are usable because the diagonal-covariance assumption keeps per-iteration cost linear in the feature dimension.

Reading between the lines

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

  • An extension the authors do not pursue: the same empirical-Fisher scaling applies to any exponential-family variational posterior with computable gradients, so topic models and other Bayesian nonparametrics are natural next targets.
  • A test the paper does not run: on a synthetic Gaussian DPM with known posterior covariance, SGA+F's effective per-coordinate stepsize should track the inverse posterior variance; if it does not, the natural-gradient interpretation fails.
  • Because $F_\theta$ is recomputed from each random minibatch, the adaptive stepsize inherits minibatch noise; the class-balanced minibatches used in all experiments are probably a silent condition for the reported stability.
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

4 major / 7 minor

Summary. The manuscript proposes stochastic gradient ascent (SGA) updates for variational inference in Dirichlet process mixture models, with two variants: SGA+M (momentum) and SGA+F (Fisher-information-scaled adaptive stepsize). The authors claim that SGA learning is faster than closed-form coordinate ascent (the MM baseline) without sacrificing NMI or accuracy, that the Fisher information automatically provides an optimal stepsize, and that the approach scales to large datasets (Caltech256, SUN397) with deep ConvNet features. The paper reports experiments on six object/scene datasets, including model selection, CPU time, NMI, and accuracy, and compares with several Bayesian nonparametric baselines from the literature.

Significance. If the central claims were established, the paper would offer a practical closed-form-free DPM inference algorithm with adaptive stepsize, applicable to high-dimensional features and datasets with up to 397 classes. The authors provide pseudocode and a convergence diagnostic based on the variational lower bound. However, the theoretical justification for the Fisher-information stepsize is not valid as written, and the experimental protocol contains confounds that prevent the reported improvements from being attributed to the proposed method. With the derivation corrected and the method reframed as a heuristic adaptive-gradient preconditioner, an empirical contribution might remain, but the current manuscript does not support the 'automatic optimal stepsize' claim or the comparative performance claims.

major comments (4)
  1. [Section 3.3, Eqs. (10)-(13)] The natural-gradient derivation is misapplied. The Fisher information matrix in natural gradient is the expected outer product of the gradient of log q(theta; phi) with respect to the variational parameters phi. Here q is the variational posterior over hidden variables theta = (mu, v), and Eq. (10) defines F_theta as the expectation of squared gradients with respect to theta itself. For a Gaussian q(mu) with fixed variance, the Fisher information with respect to the mean is a constant, whereas Eq. (11) is data-dependent; if q is a point-mass MAP estimate, the Fisher metric is not even well-defined. Consequently, Eq. (13) does not follow from the KL-constrained optimization in Eq. (12), and the abstract's 'automatic optimal stepsize' is not established. At best, Eq. (9) is a diagonal adaptive-gradient preconditioner reminiscent of RMSprop, not a natural-gradient update.
  2. [Section 5.2, Table 2] The claim that 'all SGAs outperforms MM on all 5 evaluation criteria' is contradicted by Table 2. For Caltech10, the SGA+M model estimate is 24 versus MM's 13.3 (ground truth 10); for Caltech101, SGA+M is 139.7 versus MM's 126.3; for Scene15, SGA+F is 18 versus MM's 17.7. Thus not even the SGA+F variant is uniformly closer to ground truth across all datasets, and the SGA+M variant is often worse than MM. The summary in the text overstates the empirical evidence.
  3. [Sections 5.2 and 7.2, Table 3] The comparison against the MM baseline is not a fair test of closed-form coordinate ascent. For the larger datasets, MM is restricted to a fixed random subset of about 5K images ('there is a large portion of dataset that MM will never see'), while the SGA methods see the entire dataset through random minibatches. The reported CPU-time and accuracy gains therefore conflate algorithmic advantage with access to more data. In addition, the feature extractor is inconsistent: Section 5.1.4 states VGG16 pretrained on ImageNet, while Section 7.2 states that the proposed method uses VGG16 pretrained on Place205. Table 3 also compares methods with different feature types (SIFT, Fisher vector, AlexNet FC7, VGG16), so the NMI/ACC differences cannot be attributed to the DPM learner.
  4. [Sections 3.1 and 4.3] The paper's title promises theory, but no convergence theorem is provided for Algorithm 1. Eq. (7) cites a Barzilai-Borwein stepsize for stochastic gradient descent, but that rule is never used in the proposed algorithm. Section 4.3 only checks pointwise concavity of coordinate-wise log-posteriors, which does not establish convergence of the alternating stochastic updates. The paper should either provide a formal convergence analysis or remove the theoretical claim from the title and abstract.
minor comments (7)
  1. [Abstract and Section 1] There are numerous typos and grammatical errors, including 'As as intermediate approach', 'do not sacrifice performance for speed', 'motiviation', 'datset', 'trunction level', and 'reflex'. The manuscript needs thorough proofreading.
  2. [Section 5.2 and Algorithm 1] Section 5.2 states that all SGA learners use a fixed learning rate eta = 0.1, while Algorithm 1 line 3(i) gives the initialization range 1e-1 <= eta <= 1e-3. Clarify how eta is chosen and whether any sensitivity analysis was performed.
  3. [Section 7.5] The text references 'Table 6' when discussing LDPO-A-FC, but the manuscript contains only Table 3; correct the cross-reference.
  4. [Section 5.1.2] The equations for Accuracy and NMI are typeset incorrectly ('Uinf o', 'Hrefers'), making the definitions hard to read; rewrite them cleanly.
  5. [Section 3.3] There is a typo 'eqn (121)' in the discussion of Eq. (12), and the statement that F_theta is a scalar under diagonal covariance should be clarified: the diagonal entries of the Fisher matrix are used, not a scalar Fisher information.
  6. [Tables 2 and 3] No error bars or variance estimates are reported for NMI/ACC despite the statement that experiments were averaged over at least 5 runs; adding standard deviations would improve interpretability.
  7. [Section 4.5] The pruning threshold T_HR and the re-ordering of E[v_k] are described heuristically; please specify how T_HR was set for each dataset and whether the re-ordering affects model selection.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: the adaptive stepsize is an input normalization, the comparison baseline is re-run, and the central performance claims are empirical.

full rationale

The paper's load-bearing claims are empirical comparisons against the MM baseline and literature results, not predictions derived from fitted constants. SGA+F's Fisher information F_theta is defined in Eq. (10) as the empirical second moment of the gradient, and Eq. (13) is a standard natural-gradient/KL-constraint result cited from Martens [32]; whether this is the correct curvature metric is a modeling/correctness issue, not circularity, because the scaling is part of the update rule rather than a parameter fitted to the evaluation metric. The MM baseline [28] is a self-citation (shared first author), but Section 5.2 states 'We rerun the experiments for at least 5 times and take their average results for each dataset,' so the comparison does not depend on numbers imported from [28]; the convergence-of-MM quote from [28] supports the baseline, not the proposed SGA+F contribution. No fitted input is renamed as a prediction: minibatch sizes are empirically adjusted, but this is hyperparameter tuning rather than a prediction of NMI/accuracy. The Table 3 feature mismatch (Place205 in Section 7.2 versus ImageNet in Section 5.1.4) is an experimental confound and a correctness concern, not a circular derivation. The SGA+F update resembles RMSprop/Adagrad, which the paper acknowledges, so treating it as a novel Fisher-information derivation is a novelty/correctness concern, not circularity. Overall the derivation chain is self-contained: the gradients in Eqs. (16) and (18) are computed from the model, the Fisher scaling is an input preconditioner, and the empirical claims are tested against external and re-run baselines.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The central claims rest on standard VI machinery plus a heuristic adaptive-gradient scaling. The hand-set parameters (η, minibatch size, threshold T_HR, truncation K) materially affect the reported model selection and performance numbers, and their values are not reported fully.

free parameters (6)
  • Learning rate η = 0.1
    Fixed at 0.1 in all experiments despite the abstract's claim of automatic stepsize; the initialization range in Algorithm 1 is ambiguous.
  • Momentum coefficient α = 0.9
    Standard fixed value for the SGA+M variant; not tuned.
  • Minibatch size = 100-300 for Caltech10/Scene15; 1005-2560 for larger sets
    Empirically adjusted per dataset to achieve on-par performance with MM (Section 5.2).
  • Pruning threshold T_HR = not specified
    Section 4.5 states 'a typical setting is to use T_HR > 0' but no concrete value is given; the model selection results depend on this choice.
  • Truncation level K = 50, 100, 200, 500, 1000 (per dataset)
    Chosen as roughly 2-2.5x the ground truth for larger datasets; the estimated cluster counts in Table 2 depend heavily on K and the threshold.
  • Prior hyperparameters λ0, m0, σ, a0 = Not reported; a0 set to N in Algorithm 1
    Stated as constants but values are unspecified; clustering results are sensitive to them.
assumptions (5)
  • standard math Mean-field factorization q(θ) = ∏_i q(θ_i)
    Used throughout (Eq. 3), standard variational inference assumption.
  • domain assumption Gaussian mixture with diagonal covariance Σ_k = σ²_k I and constant σ_k = σ
    Section 4.4 states this assumption, enabling high-dimensional VGG16 features.
  • domain assumption Truncated stick-breaking representation of the Dirichlet process
    Section 2 relies on [7]'s truncated stick-breaking to set a finite truncation level K.
  • standard math Concavity of variational log-posteriors
    Section 4.3 verifies ∇² ln q ≤ 0 for μ and v under the model assumptions, used to argue SGA convergence.
  • domain assumption Ground-truth labels of benchmark datasets are correct
    Required for NMI and accuracy evaluation; standard benchmark assumption.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Fisher Information based Stochastic Gradient Ascent for Online Learning of Dirichlet Process Mixture and Theory." pith.science (2026). https://pith.science/paper/7AGXC4UQ

@misc{pith2026241208951,
  author       = {Pith},
  title        = {Pith review of: Fisher Information based Stochastic Gradient Ascent for Online Learning of Dirichlet Process Mixture and Theory},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/7AGXC4UQ}},
  note         = {Machine review of arXiv:2412.08951}
}
read the original abstract

Scalable algorithms of posterior approximation allow Bayesian nonparametrics such as Dirichlet process mixture to scale up to larger dataset at fractional cost. Recent algorithms, notably the stochastic variational inference performs local learning from minibatch. The main problem with stochastic variational inference is that it relies on closed form solution. Stochastic gradient ascent is a modern approach to machine learning and is widely deployed in the training of deep neural networks. In this work, we explore using stochastic gradient ascent as a fast algorithm for the posterior approximation of Dirichlet process mixture. However, stochastic gradient ascent alone is not optimal for learning. In order to achieve both speed and performance, we turn our focus to stepsize optimization in stochastic gradient ascent. As as intermediate approach, we first optimize stepsize using the momentum method. Finally, we introduce Fisher information to allow adaptive stepsize in our posterior approximation. In the experiments, we justify that our approach using stochastic gradient ascent do not sacrifice performance for speed when compared to closed form coordinate ascent learning on these datasets. Lastly, our approach is also compatible with deep ConvNet features as well as scalable to large class datasets such as Caltech256 and SUN397.

Figures

Figures reproduced from arXiv: 2412.08951 by the authors.

Figure 1
Figure 1. CPU time (as a factor of MM) and minibatch size [PITH_FULL_IMAGE:figures/full_fig_p023_1.png] view at source ↗
Figure 2
Figure 2. NMI and Accuracy 24 [PITH_FULL_IMAGE:figures/full_fig_p024_2.png] view at source ↗
Figure 3
Figure 3. Convergence plots of SGA (top) and SGA+Fisher (bot) on Caltech256 [PITH_FULL_IMAGE:figures/full_fig_p026_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Convergence plots of SGA (top) and SGA+Fisher (bot) on SUN397 [PITH_FULL_IMAGE:figures/full_fig_p027_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

39 extracted references · 37 canonical work pages

  1. [1]

    Zhang, L

    W. Zhang, L. Du, L. Li, X. Zhang, H. Liu, Infinite bayesian one-class sup- port vector machine based on dirichlet process mixture clustering, Pattern Recognition 78 (2018) 56–78

    work page 2018

  2. [2]

    H. Wu, S. Prasad, Semi-supervised dimensionality reduction of hyperspec- tral imagery using pseudo-labels, Pattern Recognition 74 (2018) 212–224

    work page 2018

  3. [3]

    X. Ye, J. Zhao, Multi-manifold clustering: A graph-constrained deep non- parametric method, Pattern Recognition 93 (2019) 215–227

    work page 2019

  4. [4]

    Z. Ma, P. K. Rana, J. Taghia, M. Flierl, A. Leijon, Bayesian estimation of dirichlet mixture model with variational inference, Pattern Recognition 47 (9) (2014) 3143–3157

    work page 2014

  5. [5]

    W. Fan, N. Bouguila, Variational learning of a dirichlet process of general- ized dirichlet distributions for simultaneous clustering and feature selection, Pattern Recognition 46 (10) (2013) 2754–2769

    work page 2013

  6. [6]

    Liu, H.-C

    C. Liu, H.-C. Li, K. Fu, F. Zhang, M. Datcu, W. J. Emery, Bayesian estimation of generalized gamma mixture model based on variational em algorithm, Pattern Recognition 87 (2019) 269–284

    work page 2019

  7. [7]

    D. M. Blei, M. I. Jordan, et al., Variational inference for dirichlet process mixtures, Bayesian analysis 1 (1) (2006) 121–144. 31

    work page 2006

  8. [8]

    M. D. Hoffman, D. M. Blei, C. Wang, J. W. Paisley, Stochastic variational inference., Journal of Machine Learning Research 14 (1) (2013) 1303–1347

    work page 2013

Show all 39 references
  1. [9]

    Mandt, M

    S. Mandt, M. D. Hoffman, D. M. Blei, Stochastic gradient descent as ap- proximate bayesian inference, The Journal of Machine Learning Research 18 (1) (2017) 4873–4907

    2017

  2. [10]

    Ranganath, S

    R. Ranganath, S. Gerrish, D. Blei, Black box variational inference, in: Artificial Intelligence and Statistics, 2014, pp. 814–822

    2014

  3. [11]

    Paisley, D

    J. Paisley, D. M. Blei, M. I. Jordan, Variational bayesian inference with stochastic search, in: Proceedings of the 29th International Coference on International Conference on Machine Learning, Omnipress, 2012, pp. 1363– 1370

    2012

  4. [12]

    D. P. Kingma, M. Welling, Stochastic gradient vb and the variational auto- encoder, in: Second International Conference on Learning Representations, ICLR, 2014

    2014

  5. [13]

    Welling, Y

    M. Welling, Y. W. Teh, Bayesian learning via stochastic gradient langevin dynamics, in: Proceedings of the 28th International Conference on Machine Learning (ICML-11), 2011, pp. 681–688

    2011

  6. [14]

    D. J. Rezende, S. Mohamed, Variational inference with normalizing flows, in: Proceedings of the 32nd International Conference on International Con- ference on Machine Learning-Volume 37, JMLR. org, 2015, pp. 1530–1538

    2015

  7. [15]

    Robbins, S

    H. Robbins, S. Monro, A stochastic approximation method, in: Herbert Robbins Selected Papers, Springer, 1985, pp. 102–109

    1985

  8. [16]

    C. Tan, S. Ma, Y.-H. Dai, Y. Qian, Barzilai-borwein step size for stochastic gradient descent, in: Advances in Neural Information Processing Systems, 2016, pp. 685–693. 32

    2016

  9. [17]

    Honkela, T

    A. Honkela, T. Raiko, M. Kuusela, M. Tornio, J. Karhunen, Approximate riemannian conjugate gradient learning for fixed-form variational bayes, Journal of Machine Learning Research 11 (Nov) (2010) 3235–3268

    2010

  10. [18]

    Duchi, E

    J. Duchi, E. Hazan, Y. Singer, Adaptive subgradient methods for online learning and stochastic optimization, Journal of Machine Learning Re- search 12 (Jul) (2011) 2121–2159

    2011

  11. [19]

    D. P. Kingma, J. Ba, Adam: A method for stochastic optimization, arXiv preprint arXiv:1412.6980 (2014)

    2014 arXiv

  12. [20]

    Kulis, M

    B. Kulis, M. I. Jordan, Revisiting k-means: New algorithms via bayesian nonparametrics, in: Proceedings of the 29th International Conference on Machine Learning (ICML-12), 2012, pp. 513–520

    2012

  13. [21]

    Broderick, N

    T. Broderick, N. Boyd, A. Wibisono, A. C. Wilson, M. I. Jordan, Streaming variational bayes, in: Advances in neural information processing systems, 2013, pp. 1727–1735

    2013

  14. [22]

    Kurihara, M

    K. Kurihara, M. Welling, Bayesian k-means as a maximization-expectation algorithm, Neural computation 21 (4) (2009) 1145–1172

    2009

  15. [23]

    Sato, Online model selection based on the variational bayes, Neural computation 13 (7) (2001) 1649–1681

    M.-A. Sato, Online model selection based on the variational bayes, Neural computation 13 (7) (2001) 1649–1681

    2001

  16. [24]

    W. Fan, H. Sallay, N. Bouguila, Online learning of hierarchical pitman–yor process mixture of generalized dirichlet distributions with feature selection, IEEE transactions on neural networks and learning systems 28 (9) (2016) 2048–2061

    2016

  17. [25]

    Patacchiola, A

    M. Patacchiola, A. Cangelosi, Head pose estimation in the wild using con- volutional neural networks and adaptive gradient methods, Pattern Recog- nition 71 (2017) 132–143. 33

    2017

  18. [26]

    F. Li, H. Qiao, B. Zhang, Discriminatively boosted image clustering with fully convolutional auto-encoders, Pattern Recognition 83 (2018) 161–173

    2018

  19. [27]

    C. M. Bishop, Pattern recognition and machine learning, springer, 2006

    2006

  20. [28]

    K.-L. Lim, H. Wang, Fast approximation of variational bayes dirichlet process mixture using the maximization–maximization algorithm, Inter- national Journal of Approximate Reasoning 93 (2018) 153–177

    2018

  21. [29]

    R. M. Neal, G. E. Hinton, A view of the em algorithm that justifies in- cremental, sparse, and other variants, in: Learning in graphical models, Springer, 1998, pp. 355–368

    1998

  22. [30]

    D. M. Titterington, The em algorithm, variational approximations and ex- pectationpropagationformixtures, Mixtures: EstimationandApplications 896 (2011)

    2011

  23. [31]

    Honkela, M

    A. Honkela, M. Tornio, T. Raiko, J. Karhunen, Natural conjugate gradient in variational inference, in: International Conference on Neural Information Processing, Springer, 2007, pp. 305–314

    2007

  24. [32]

    Martens, New insights and perspectives on the natural gradient method, arXiv preprint arXiv:1412.1193 (2014)

    J. Martens, New insights and perspectives on the natural gradient method, arXiv preprint arXiv:1412.1193 (2014)

    2014 arXiv

  25. [33]

    Ç. Arı, S. Aksoy, O. Arıkan, Maximum likelihood estimation of gaussian mixture models using stochastic search, Pattern Recognition 45 (7) (2012) 2804–2816

    2012

  26. [34]

    Kurihara, M

    K. Kurihara, M. Welling, Y. W. Teh, Collapsed variational dirichlet process mixture models., in: IJCAI, Vol. 7, 2007, pp. 2796–2801

    2007

  27. [35]

    D. Cai, X. He, J. Han, Document clustering using locality preserving in- dexing, IEEE Transactions on Knowledge and Data Engineering 17 (12) (2005) 1624–1637. 34

    2005

  28. [36]

    C. Chen, J. Zhu, X. Zhang, Robust bayesian max-margin clustering, in: Advances in Neural Information Processing Systems, 2014, pp. 532–540

    2014

  29. [37]

    Nguyen, D

    V. Nguyen, D. Phung, T. Le, H. Bui, Discriminative bayesian nonparamet- ric clustering, in: IJCAI 2017: Proceedings of the 26th International Joint Conference on Artificial Intelligence, AAAI Press, 2017, pp. 2550–2556

    2017

  30. [38]

    X. Wang, L. Lu, H.-C. Shin, L. Kim, M. Bagheri, I. Nogues, J. Yao, R. M. Summers, Unsupervised joint mining of deep features and image labels for large-scale radiology image categorization and scene recognition, in: Applications of Computer Vision (WACV), 2017 IEEE Winter Conf...

    2017

  31. [39]

    W. Fan, H. Sallay, N. Bouguila, S. Bourouis, Variational learning of hier- archical infinite generalized dirichlet mixture models and applications, Soft Computing 20 (3) (2016) 979–990. 35

    2016

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.