Pith. sign in

REVIEW 3 major objections 4 minor 54 references

Structured Basis Function Networks: Loss-Centric Multi-Hypothesis Ensembles with Controllable Diversity

T0 review · 3 major / 4 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read Structured Basis Function Networks claim to unify multi-hypothesis prediction and ensemble learning by aggregating base predictors at the Bregman centroid of the task loss, with a tunable diversity parameter that controls the bias-variance-

desk verdict The method is a reasonable incremental extension, but the identical MNIST/CIFAR-10 numbers in Tables 7–10 make the empirical core unusable as printed. read the letter →

arxiv 2509.02792 v1 pith:JLAKFJ4R submitted 2025-09-02 cs.LG

Alejandro Rodriguez Dominguez , Muhammad Shahzad , Xia Hong This is my paper
classification cs.LG
keywords Bregmandivergencesdeepensemblesmulti-hypothesispredictionstructuredbasisfunctionnetworksdiversitycontrolcentroidalaggregationbias-variance-diversitytrade-offlossgeometry
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 argues that multi-hypothesis prediction and ensemble learning are two halves of one problem: how to combine several plausible outputs in a way that respects the geometry of the loss being optimized. Its central object is the Structured Basis Function Network (s-BFN), a combiner that maps the concatenated predictions of base models through radial-basis features and linearly aggregates them, with the loss chosen by the user (squared loss for regression, cross-entropy for classification). The paper claims this s-BFN output is the Bregman centroid of the base predictions, so that the combiner is aligned with the loss geometry rather than a generic averaging rule. A single scalar parameter epsilon reweights the base learners' updates, interpolating between hard winner-takes-all training and uniform updates, and thereby giving parametric control over diversity. Experiments across regression and image classification report that s-BFN consistently outperforms logit averaging, Mixture-of-Experts, and arithmetic-mean baselines, with the best results at intermediate epsilon and heterogeneous base architectures.

What carries the argument

The load-bearing object is the structured basis function network (s-BFN): a radial-basis feature map Phi(D_i; theta) = [phi(||D_i - C_k||/gamma_k)] over the vector of all base predictions, followed by a linear head alpha (and a temperature-scaled softmax in classification). This map is what the paper identifies with the Bregman centroid z_hat = argmin_z sum_j alpha_j B_phi(z_j, z); the feature centers are running statistics of base predictions, and alpha is fit by least squares or gradient descent. The diversity mechanism is the epsilon-modulated update delta_i^(j) = 1-epsilon for the best predictor and epsilon/(M-1) for the rest, which controls how winner-takes-all the training is and thus

What would settle it

On a regression problem with M known base predictors, compute the exact Bregman centroid for squared loss - the weighted arithmetic mean sum_j alpha_j f_j / sum_j alpha_j - and compare it with the s-BFN output Phi(D_i) alpha on a held-out set. If the two disagree on a material fraction of inputs (beyond optimization tolerance), the claim that s-BFN implements centroidal aggregation is falsified; likewise for cross-entropy, compare to the KL centroid of the base probability vectors.

Watch

Extended reading notes

Core claim

The paper's central claim is that a combiner built as a structured basis function network over base-prediction vectors coincides with the loss-induced Bregman centroid, so that the same framework covers regression and classification by choosing the matching divergence. For squared loss, the combiner is a linear readout of radial-basis features over the concatenated scalars; for cross-entropy, it is a temperature-scaled softmax over radial-basis features of the base models' probability vectors. The combiner weights admit a closed-form ridge solution in the Euclidean case and gradient-based training in general, and a diversity parameter epsilon modulates each base learner's update so that the

Load-bearing premise

The paper assumes that a radial-basis function whose centers are running statistics of base predictions can actually represent the loss-induced Bregman centroid, so that training the combiner with the task loss gives a truly loss-aware ensemble; if that representational assumption fails, the combiner still minimizes training loss but the centroidal interpretation is unsupported.

Editorial extensions

If this is right

  • If the central claim holds, multi-hypothesis prediction and ensembling share one principled combination rule: the Bregman centroid of the loss, rather than ad hoc averaging or gating.
  • Practitioners get a direct knob over the bias-variance-diversity trade-off, with evidence that intermediate diversity (epsilon in roughly [0.3, 0.8]) and heterogeneous base architectures give the best accuracy.
  • The closed-form least-squares solution for squared losses and gradient-based training for arbitrary losses means the same framework can be extended to non-Euclidean losses such as Itakura-Saito divergence.
  • The reported ranking of combiners - s-BFN above logit averaging, MoE, and arithmetic mean - implies that learned probability-space aggregation can be a robust default for deep ensembles, especially when base models are miscalibrated.
  • Optimal ensemble size depends on dataset complexity and diversity, so capacity should be scaled jointly with diversity rather than treated as an independent hyperparameter.

Reading between the lines

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

  • An extension the paper leaves implicit: if the centroidal identification is taken literally, s-BFN should reproduce the exact Bregman centroid as the number of basis functions and training data grow; one testable prediction is that accuracy gains concentrate where base predictions are well separated into Voronoi cells.
  • The epsilon mechanism suggests a continuous family of training objectives between hard winner-takes-all and uniform updates; scheduling epsilon over training or per hypothesis could adapt diversity to local data complexity more flexibly than a global constant.
  • Because the combiner feeds on softmax probabilities rather than logits, its advantage over logit averaging is likely most pronounced when base softmax outputs are miscalibrated; temperature-scaling the base probabilities before the RBF map is a natural intervention the paper does not explore.
  • The representational claim that RBF features with running-statistic centers can express the loss centroid is the fragile link; a parameter-free diagnostic comparing s-BFN outputs to closed-form centroids on held-out base predictions would directly probe it.
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

3 major / 4 minor

Summary. The paper introduces Structured Basis Function Networks (s-BFN), a combiner that maps a concatenation of base-predictor outputs through a radial-basis feature map and a linear readout, with the claimed property that the resulting prediction is the Bregman centroid induced by the task loss (squared loss for regression, cross-entropy for classification). A tunable diversity parameter ε modulates base-learner updates, and the framework admits both a closed-form ridge solution and gradient-based training. Experiments on tabular regression and MNIST/Fashion-MNIST/CIFAR-10 image classification are presented to support claims that s-BFN outperforms logit averaging, Mixture-of-Experts, and arithmetic-mean baselines, and to study a complexity–capacity–diversity trade-off.

Significance. If the central claims held, the paper would offer a useful synthesis of multi-hypothesis prediction and loss-geometry-aware ensembling. The closed-form ridge estimator (Eq. 7) and the gradient updates in Section 3.3.2 are standard and internally consistent, and the explicit diversity-control mechanism is a practical idea. However, the empirical core is compromised by an internal inconsistency in the reported tables, and the theoretical link between the s-BFN output and the Bregman centroid is asserted rather than proven. As printed, the paper does not support its headline empirical claims.

major comments (3)
  1. [Tables 7–10 (Section 4.5)] Tables 7, 8, 9, and 10 report exactly identical values for MNIST and CIFAR-10: Table 7 gives the same sRBF/MoE/Base accuracies (49.10/47.96/39.71 and 32.70/33.45/30.38), Table 8 gives identical best-configuration accuracies, Table 9 gives identical train/eval times, FLOPs, and parameters, and Table 10 gives identical sensitivity values for every ε. MNIST and CIFAR-10 are different datasets, so exact equality to two decimals across all metrics cannot occur by chance. This indicates a copy/paste or labeling error. Section 4.4's central claim that 's-BFN achieves the highest accuracy, followed by logit averaging, MoE, and arithmetic-mean' across datasets is therefore unsupported by the printed evidence. This is a load-bearing error: the main empirical contribution cannot be verified without corrected tables or the underlying raw results/code.
  2. [Section 3.2.1, Figure 1] The paper asserts that the s-BFN output Φ(Di)α is the Bregman centroid of the base predictions, but no derivation or theorem establishes this. For cross-entropy, the Bregman centroid under KL divergence is a weighted geometric mean of probability vectors (equivalently, softmax of weighted logits). The s-BFN instead applies a radial-basis feature map to concatenated probability vectors and a linear readout; even if α is trained with the task loss, this does not imply that the resulting prediction equals the centroid. The RBF feature map may not be expressive enough, and no convergence or representational result is given. This matters because 'centroidal aggregation' is a central claim of the abstract and introduction. At minimum, the paper should state this as an approximation and provide a concrete test (e.g., synthetic mixtures where the true centroid is known) or a proof under explicit
  3. [Section 4.4 and Figures 3–8] Several quantitative claims about the complexity–capacity–diversity trade-off are based on figures that are described but whose numerical values are not reported in tables. For instance, the claims that 'accuracy peaks at moderate M on MNIST' and 'on CIFAR-10, capacity gains materialize only when diversity is sufficiently high' are asserted from plots without accompanying summary statistics or error bars in the text. Given that the only dataset-level tables (Tables 7–10) contain the MNIST/CIFAR-10 duplication, these figure-dependent conclusions are not independently checkable from the manuscript. Please provide per-configuration numerical results or a data/code release.
minor comments (4)
  1. [Algorithm 1] The regularization term in the update is written inconsistently: Eq. (8) uses λ1^j ∇θj Rθj(θj), while Algorithm 1 writes λ1^j/N sgn(θj). Please align the notation and specify the exact regularization used.
  2. [Section 3.2.1] The phrase 'the ensemble output must follow the associated centroidal rule' is stated as fact. Since the Bregman centroid is the optimal combiner for a given loss only under a specific aggregation model, please qualify this statement and cite the precise assumptions from Wood et al. (2024).
  3. [Various] There is no code or data availability statement. Given the empirical inconsistencies, providing code or raw output tables would be essential for reproducibility.
  4. [Appendix A] The comparison between GBF and RBF-KMeans is useful, but the description of center initialization says GBF centers are 'uniformly assigned' while RBF-KMeans centers are 'derived from unsupervised clustering.' Please clarify whether GBF centers are fixed or learned, and how this affects the claimed advantages.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found: the s-BFN combiner is trained by direct loss minimization (ridge or gradient), and the Bregman-centroid claim, while under-derived, is an assertion rather than an equation that reduces to its own input.

full rationale

Walking the derivation chain, Section 3.3 defines the s-BFN output as Φ(D_i)α and Eq. (7) is the standard ridge-regression solution; Algorithms 1–2 minimize MSE and cross-entropy respectively. The paper asserts in Section 3.2.1 and Figure 1 that the s-BFN output is the loss-induced Bregman centroid, but training α by the task loss is not, by construction, equivalent to solving ẑ = argmin_z Σ_j α_j Bφ(z_j, z). That is a missing derivation / support gap, not a circular reduction: no equation is defined in terms of the result it is used to predict, and no fitted parameter is renamed as a prediction. The self-citation to Dominguez et al. (2025) for the closed-form estimator is not load-bearing because Eq. (7) is derived in the present text; the Bregman-centroid background is cited from Wood et al. (2024), which is not a self-citation. I also flag that Tables 7–10 report identical MNIST and CIFAR-10 numbers, which is an empirical/internal-consistency problem, but it is a correctness issue rather than circularity. No step in the paper reduces its stated prediction to the data or fitted values by definition, so the circularity score is 0.

Assumptions & free parameters 7 free parameters · 4 assumptions · 0 invented entities

The central result depends on external Bregman-divergence theory, a heuristic RBF feature representation, and an empirically chosen diversity schedule. No theorem connects the learned alpha to the claimed centroid, and the epsilon values that deliver the reported gains are selected after seeing test performance.

free parameters (7)
  • diversity parameter epsilon = 0.35 best on tabular regression; 0.5-0.8 best on image classification
    Central knob of the method. Searched over {0,0.1,0.35,0.5} in Table 3 and over {0,0.3,0.5,0.8} in image experiments; the best value is selected after seeing test performance.
  • number of hypotheses/base learners M = 2,5,10,20,35 searched; best configurations use M=5 or 10
    Ensemble size is tuned per dataset and directly controls the dimension of the structured dataset.
  • RBF centers C and scales gamma = computed from running statistics (G1-G3) or updated by gradient (Algorithm 2)
    The feature map parameters are data-dependent and heuristic; their values determine the effective basis and hence the combiner's behaviour.
  • regularization lambda1, lambda2 = lambda1 in {0,0.0001,0.01,0.07}; lambda2 in {0,3,5,7}
    Tuned per dataset; the closed-form solution in Eq. (7) depends directly on lambda2.
  • learning rates eta_theta, eta_alpha = 0.03, 0.3, 1.0 searched; best classification runs use eta_alpha=0.1
    All gradient updates depend on these rates; Table 8 reports the best alpha learning rate.
  • temperature T = 3 for best configurations
    Scales the combiner logits before the final softmax in classification (Algorithm 2, Table 8).
  • number of RBF units K = K=M for regression; K can be M*k for classification, k unspecified
    The basis-function count is a capacity parameter; the exact K used in the image experiments is not reported.
assumptions (4)
  • domain assumption The optimal way to combine base predictions for a given loss is the Bregman centroid of those predictions (Wood et al., 2024).
    The paper imports this result and treats it as the foundation for its combiner, without independent proof in this manuscript.
  • ad hoc to paper A radial-basis feature map with centers and scales computed from running statistics can represent the Bregman centroid sufficiently well.
    No approximation guarantee is given; the centers/scales are heuristic (G1-G3, Algorithm 2), so the claimed loss-alignment depends on this.
  • domain assumption Base predictors trained with the epsilon-modulated update in Eq. (9) will specialize in the desired way without mode collapse.
    The paper relies on the WTA-style update plus nonzero updates for losers to produce complementary hypotheses; this is validated only empirically.
  • domain assumption The hyperparameter grid (Table 3) and holdout splits provide unbiased estimates of generalization.
    The paper reports best configurations from 80 hyperparameter combinations and 5 splits, but does not describe a separate validation procedure for the classification experiments.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Structured Basis Function Networks: Loss-Centric Multi-Hypothesis Ensembles with Controllable Diversity." pith.science (2026). https://pith.science/paper/JLAKFJ4R

@misc{pith2026250902792,
  author       = {Pith},
  title        = {Pith review of: Structured Basis Function Networks: Loss-Centric Multi-Hypothesis Ensembles with Controllable Diversity},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/JLAKFJ4R}},
  note         = {Machine review of arXiv:2509.02792}
}
read the original abstract

Existing approaches to predictive uncertainty rely either on multi-hypothesis prediction, which promotes diversity but lacks principled aggregation, or on ensemble learning, which improves accuracy but rarely captures the structured ambiguity. This implicitly means that a unified framework consistent with the loss geometry remains absent. The Structured Basis Function Network addresses this gap by linking multi-hypothesis prediction and ensembling through centroidal aggregation induced by Bregman divergences. The formulation applies across regression and classification by aligning predictions with the geometry of the loss, and supports both a closed-form least-squares estimator and a gradient-based procedure for general objectives. A tunable diversity mechanism provides parametric control of the bias-variance-diversity trade-off, connecting multi-hypothesis generalisation with loss-aware ensemble aggregation. Experiments validate this relation and use the mechanism to study the complexity-capacity-diversity trade-off across datasets of increasing difficulty with deep-learning predictors.

Figures

Figures reproduced from arXiv: 2509.02792 by the authors.

Figure 1
Figure 1. Workflow: base predictors produce a structured dataset [PITH_FULL_IMAGE:figures/full_fig_p011_1.png] view at source ↗
Figure 2
Figure 2. Mean RMSE and 90% confidence interval from 10-fold cross-validation for different configurations of hyper-parameters M and ε. 15 [PITH_FULL_IMAGE:figures/full_fig_p015_2.png] view at source ↗
Figure 3
Figure 3. Results on the MNIST dataset across different ensemble settings. [PITH_FULL_IMAGE:figures/full_fig_p019_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Results on the Fashion-MNIST dataset across different ensemble settings. [PITH_FULL_IMAGE:figures/full_fig_p020_4.png]
Figure 5
Figure 5. Figure 5: Results on CIFAR-10 [PITH_FULL_IMAGE:figures/full_fig_p021_5.png]
Figure 6
Figure 6. Figure 6: Gibbs risk RG(ε) and expected disagreement d(ε) (left axis), together with majority-vote error RMV (ε) and its PAC-Bayes upper bound (right axis), for varying diversity parameter ε. The curves illustrate an intermediate ε where the bound and empirical error are minimiz…
Figure 7
Figure 7. Figure 7: Computational efficiency evaluation for different ensemble configurations on CIFAR-10. [PITH_FULL_IMAGE:figures/full_fig_p022_7.png]
Figure 8
Figure 8. Figure 8: Computational efficiency results on Fashion-MNIST. [PITH_FULL_IMAGE:figures/full_fig_p023_8.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

54 extracted references · 45 canonical work pages

  1. [1]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in ":" * " " * FUNCTION f...

  2. [2]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in ":" * " " * FUNCTION f...

  3. [3]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in ":" * " " * FUNCTION f...

  4. [4]

    Pathologies of Predictive Diversity in Deep Ensembles

    Abe, T., Buchanan, E. K., Pleiss, G., and Cunningham, J. P. (2023). Pathologies of predictive diversity in deep ensembles. CoRR , abs/2302.00704

    work page Pith review arXiv 2023

  5. [5]

    Arghira, N., Hawarah, L., Ploix, S., and Jacomino, M. (2012). Prediction of appliances energy use in smart homes. Energy , 48(1):128--134. 6th Dubrovnik Conference on Sustainable Development of Energy Water and Environmental Systems, SDEWES 2011

    work page 2012

  6. [6]

    Bishop, C. M. (1995). Neural networks for pattern recognition . Oxford

    work page 1995

  7. [7]

    Bregman, L. (1967). The relaxation method of finding the common point of convex sets and its application to the solution of problems in convex programming. USSR Computational Mathematics and Mathematical Physics , 7(3):200--217

    work page 1967

  8. [8]

    Breiman, L. (1996). Bagging predictors. Mach. Learn. , 24(2):123–140

    work page 1996

Show all 54 references
  1. [9]

    Brown, G. (2004). Diversity in neural network ensembles . PhD thesis, Citeseer

    2004

  2. [10]

    M., Feldheim, V., and Deramaix, D

    Candanedo, L. M., Feldheim, V., and Deramaix, D. (2017). Data driven prediction models of energy use of appliances in a low-energy house. Energy and Buildings , 140:81--97

    2017

  3. [11]

    Cetin, K., Tabares-Velasco, P., and Novoselac, A. (2014). Appliance daily energy use in new residential buildings: Use profiles and variation in time-of-use. Energy and Buildings , 84:716--726

    2014

  4. [12]

    De Vito , S., Massera, E., Piga, M., Martinotto, L., and Di Francia , G. (2008). On field calibration of an electronic nose for benzene estimation in an urban pollution monitoring scenario. Sensors and Actuators B: Chemical , 129(2):750--757

    2008

  5. [13]

    R., Shahzad, M., and Hong, X

    Dominguez, A. R., Shahzad, M., and Hong, X. (2025). Structured radial basis function network: Modelling diversity for multiple hypotheses prediction. In Bramer, M. and Stahl, F., editors, Artificial Intelligence XLI , pages 88--101, Cham. Springer Nature Switzerland

    2025

  6. [14]

    Dosovitskiy, A., Beyer, L., Kolesnikov, A., Weissenborn, D., Zhai, X., Unterthiner, T., Dehghani, M., Minderer, M., Heigold, G., Gelly, S., Uszkoreit, J., and Houlsby, N. (2021). An image is worth 16x16 words: Transformers for image recognition at scale. In International Confe...

    2021

  7. [15]

    Fort, S., Hu, H., and Lakshminarayanan, B. (2019). Deep ensembles: A loss landscape perspective. ArXiv , abs/1912.02757

    2019 arXiv

  8. [16]

    Germain, P., Lacasse, A., Laviolette, F., Marchand, M., and Roy, J.-F. (2015). Risk bounds for the majority vote: From a pac-bayesian analysis to a learning algorithm. Journal of Machine Learning Research , 16:787--860

    2015

  9. [17]

    Grandvalet, Y., d'Alch \'e Buc, F., and Ambroise, C. (2001). Boosting mixture models for semi-supervised learning. In Dorffner, G., Bischof, H., and Hornik, K., editors, Artificial Neural Networks --- ICANN 2001 , pages 41--48, Berlin, Heidelberg. Springer Berlin Heidelberg

    2001

  10. [18]

    Guzm\' a n-rivera, A., Batra, D., and Kohli, P. (2012). Multiple choice learning: Learning to produce multiple structured outputs. In Pereira, F., Burges, C., Bottou, L., and Weinberger, K., editors, Advances in Neural Information Processing Systems , volume 25. Curran Associates, Inc

    2012

  11. [19]

    Guzman-Rivera, A., Kohli, P., Batra, D., and Rutenbar, R. (2014). Efficiently Enforcing Diversity in Multi-Output Structured Prediction . In Kaski, S. and Corander, J., editors, Proceedings of the Seventeenth International Conference on Artificial Intelligence and Statistics ,...

    2014

  12. [20]

    E., Vinyals, O., and Dean, J

    Hinton, G. E., Vinyals, O., and Dean, J. (2015). Distilling the knowledge in a neural network. ArXiv , abs/1503.02531

    2015 arXiv

  13. [21]

    Ho, T. K. and Basu, M. (2002). Complexity measures of supervised classification problems. IEEE Transactions on Pattern Analysis and Machine Intelligence , 24(3):289--300

    2002

  14. [22]

    Huang, G., Li, Y., Pleiss, G., Liu, Z., Hopcroft, J., and Weinberger, K. (2017). Snapshot ensembles: Train 1, get m for free

    2017

  15. [23]

    Kondratyuk, D., Tan, M., Brown, M., and Gong, B. (2020). When ensembling smaller models is more efficient than single large models

    2020

  16. [24]

    Krizhevsky, A. (2009). Learning multiple layers of features from tiny images

    2009

  17. [25]

    and Vedelsby, J

    Krogh, A. and Vedelsby, J. (1994). Neural network ensembles, cross validation, and active learning. In Tesauro, G., Touretzky, D., and Leen, T., editors, Advances in Neural Information Processing Systems , volume 7. MIT Press

    1994

  18. [26]

    and Whitaker, C

    Kuncheva, L. and Whitaker, C. (2003). Measures of diversity in classifier ensembles and their relationship with the ensemble accuracy. Machine Learning , 51:181--207

    2003

  19. [27]

    Lakshminarayanan, B., Pritzel, A., and Blundell, C. (2017). Simple and scalable predictive uncertainty estimation using deep ensembles. In Proceedings of the 31st International Conference on Neural Information Processing Systems , NIPS'17, page 6405–6416, Red Hook, NY, USA. Cu...

    2017

  20. [28]

    Lecun, Y., Bottou, L., Bengio, Y., and Haffner, P. (1998). Gradient-based learning applied to document recognition. Proceedings of the IEEE , 86(11):2278--2324

    1998

  21. [29]

    Leemann, T., Sackmann, M., Thielecke, J., and Hofmann, U. (2021). Distribution preserving multiple hypotheses prediction for uncertainty modeling. CoRR , abs/2110.02858

    2021 arXiv

  22. [30]

    and Lee, G

    Li, C. and Lee, G. H. (2019). Generating multiple hypotheses for 3d human pose estimation with mixture density network. In 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , pages 9879--9887

    2019

  23. [31]

    Li, W., Liu, H., Tang, H., and Wang, P. (2023a). Multi-hypothesis representation learning for transformer-based 3d human pose estimation. Pattern Recognition , 141:109631

  24. [32]

    Li, Z., Ren, K., Yang, Y., Jiang, X., Yang, Y., and Li, D. (2023b). Towards inference efficient deep ensemble learning. In Proceedings of the Thirty-Seventh AAAI Conference on Artificial Intelligence and Thirty-Fifth Conference on Innovative Applications of Artificial Intellig...

  25. [33]

    Liu, X., Peng, H., Zheng, N., Yang, Y., Hu, H., and Yuan, Y. (2023). Efficientvit: Memory efficient vision transformer with cascaded group attention. In 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , pages 14420--14430

    2023

  26. [34]

    Mienye, I. D. and Sun, Y. (2022). A survey of ensemble learning: Concepts, algorithms, applications, and prospects. IEEE Access , 10:99129--99149

    2022

  27. [35]

    H., Sinthong, P., and Kalagnanam, J

    Nie, Y., Nguyen, N. H., Sinthong, P., and Kalagnanam, J. (2022). A time series is worth 64 words: Long-term forecasting with transformers. ArXiv , abs/2211.14730

    2022 arXiv

  28. [36]

    Rosset, S., Zou, H., and Hastie, T. (2006). Multi-class adaboost. Statistics and its interface , 2

    2006

  29. [37]

    Rupprecht, C., Laina, I., DiPietro, R., Baust, M., Tombari, F., Navab, N., and Hager, G. D. (2017). Learning in an uncertain world: Representing ambiguity through multiple hypotheses. In Proceedings of the IEEE International Conference on Computer Vision , pages 3591--3600

    2017

  30. [38]

    H., Khosravi, A., Alinejad-Rokny, H., Lam, H., Fuller-Tyszkiewicz, M., Acharya, U

    Shoeibi, A., Khodatars, M., Jafari, M., Ghassemi, N., Moridian, P., Alizadehsani, R., Ling, S. H., Khosravi, A., Alinejad-Rokny, H., Lam, H., Fuller-Tyszkiewicz, M., Acharya, U. R., Anderson, D., Zhang, Y., and Gorriz, J. M. (2023). Diagnosis of brain diseases in fusion of neu...

    2023

  31. [39]

    Tassi, C. R. N., Gawlikowski, J., Fitri, A. U., and Triebel, R. (2022). The impact of averaging logits over probabilities on ensembles of neural networks. In Pedroza, G., Chen, X. C., Hern \' a ndez - Orallo, J., Huang, X., Espinoza, H., Mallah, R., McDermid, J. A., and Castil...

    2022

  32. [40]

    Tax, D., Duin, R., and Breukelen, M. (1997). Comparison between product and mean classifier combination rules. In Pudil, J. and Grim, J., editors, Proc. 1st International Workshop Statistical Techniques in Pattern Recognition , pages 165--170

    1997

  33. [41]

    and GHOSH, J

    TUMER, K. and GHOSH, J. (1996). Error correlation and error reduction in ensemble classifiers. Connection Science , 8(3-4):385--404

    1996

  34. [42]

    and Nakano, R

    Ueda, N. and Nakano, R. (1996). Generalization error of ensemble estimators. In Proceedings of International Conference on Neural Networks (ICNN'96) , volume 1, pages 90--95 vol.1

    1996

  35. [43]

    Vito, S. (2008). Air Quality . UCI Machine Learning Repository. DOI : https://doi.org/10.24432/C59K5F

    2008 doi

  36. [44]

    M., Movshovitz - Attias, Y., and Eban, E

    Wang, X., Kondratyuk, D., Christiansen, E., Kitani, K. M., Movshovitz - Attias, Y., and Eban, E. (2022). Wisdom of committees: An overlooked approach to faster and more accurate models. In The Tenth International Conference on Learning Representations, ICLR 2022, Virtual Event...

    2022

  37. [45]

    Wilson, A. G. and Izmailov, P. (2020). Bayesian deep learning and a probabilistic perspective of generalization. In Larochelle, H., Ranzato, M., Hadsell, R., Balcan, M., and Lin, H., editors, Advances in Neural Information Processing Systems , volume 33, pages 4697--4708. Curr...

    2020

  38. [46]

    Wolpert, D. H. (1992). Stacked generalization. Neural Networks , 5(2):241--259

    1992

  39. [47]

    M., Reeve, H

    Wood, D., Mu, T., Webb, A. M., Reeve, H. W. J., Luj\' a n, M., and Brown, G. (2024). A unified theory of diversity in ensemble learning. J. Mach. Learn. Res. , 24(1)

    2024

  40. [48]

    Xiao, H., Rasul, K., and Vollgraf, R. (2017). Fashion-mnist: a novel image dataset for benchmarking machine learning algorithms. arXiv preprint arXiv:1708.07747

    2017 arXiv

  41. [49]

    Yuan, T., Xu, Z., Liu, B., and Tang, Y. (2025). Adaptive metric for knowledge distillation by deep bregman divergence. Neural Networks , page 108016

    2025

  42. [50]

    Zeng, A., Chen, M., Zhang, L., Xu, Q., Ma, S., and Liao, W. (2023). Are transformers effective for time series forecasting? In Proceedings of the AAAI Conference on Artificial Intelligence , volume 37, pages 11121--11128

    2023

  43. [51]

    Zhou, C., Peng, Y., Huang, W., Miao, X., Cao, Y., Wang, X., and Kong, X. (2025). A dynamic ensemble learning model for robust graph neural networks. Neural Networks , 191:107810

    2025

  44. [52]

    Zhou, H., Zhang, S., Peng, J., Zhang, S., Li, J., Xiong, H., and Zhang, W. (2022). Fedformer: Frequency enhanced decomposed transformer for long-term series forecasting. In Proceedings of the International Conference on Machine Learning (ICML) , pages 27268--27286

    2022

  45. [53]

    and Lishudi, D

    Zinchenko, S. and Lishudi, D. (2024). Star algorithm for neural network ensembling. Neural Networks , 170:364--375

    2024

  46. [54]

    N., Ali, M

    Çisil Güldoğuş, B., Abdullah, A. N., Ali, M. A., and Özöǧür Akyüz, S. (2025). Pruning the ensemble of convolutional neural networks using second-order cone programming. Neural Networks , 189:107544

    2025

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.