Pith. sign in

REVIEW 5 major objections 5 minor 68 references

Adaptive Ensemble of Classifiers with Regularization for Imbalanced Data Classification

T0 review · 5 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read Regularized ensemble beats standard classifiers on skewed data

desk verdict A new combination of resampling and weight interpolation with a clear write-up, but copied table entries and an algebraic proof error invalidate the headline claims—desk-reject this version. read the letter →

arxiv 1908.03595 v3 pith:YH7CKW22 submitted 2019-08-09 cs.LG stat.ML

Chen Wang , Chengyuan Deng , Zhoulu Yu , Dafeng Hui , Xiaofeng Gong , Ruisen Luo This is my paper · ORCID
classification cs.LGstat.ML
keywords adaptiveensembledynamicselectionimbalancedclassificationregularizationstochasticgradientdescentGaussianmixturemodelXGBoostbinary
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 proposes AER, an adaptive ensemble of classifiers with regularization, for binary classification when one class is much rarer than the other. Its central claim is that dynamic ensemble selection can be made robust to overfitting by combining two regularizers: resampling the majority class through a Gaussian mixture model to preserve global geometry, and training the ensemble weights with stochastic gradient descent, which favors minimum-norm solutions. The authors argue that this regularization improves accuracy while also lowering asymptotic time and memory complexity compared with training one classifier on the full data. On seven UCI benchmark datasets and five GMM-generated variations, they report that AER-XGBoost outperforms standard and recent baselines on most metrics, with McNemar's and Wilcoxon tests supporting the difference in most cases.

What carries the argument

The load-bearing mechanism is implicit regularization through two routes. First, Gaussian mixture model (GMM) resampling creates two types of subsets: majority-dominated views that force classifiers to respect global geometry, and nearly balanced local views that focus on the minority; Tomek links then remove noisy overlaps from the global subsets. Second, SGD training of the ensemble weights over cross-entropy loss converges to a minimum-norm solution, and the final prediction weight for each test instance interpolates the learned weight vector with the instance's log- or exponential likelihood under each Gaussian component. The complexity argument decomposes the total training cost into GMM fitting, individual classifier training, SGD, and validation, and claims that under a bound on SGD iterations the dominant term becomes $O(m^a L^{1-a} n^b)$, which is asymptotically better than the original $O(m^a n^b)$ when $a \geq 2$.

What would settle it

Check the proof's displayed identity in Section 4.1: with $t_2$ set to its upper bound $nR/L$, the SGD term $O(mnLt_2)$ becomes $O(mRn^2)$, not $O(mLt_2/R)$, so the hidden-term argument can be tested by substitution; alongside this, measure wall-clock training time of AER-XGBoost versus vanilla XGBoost as dataset size grows, since the paper itself observes that the AER version runs slower in practice.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that a dynamic ensemble can be regularized implicitly without sacrificing its adaptive power. AER fits a Gaussian mixture model to the majority class, builds 2L balanced sub-datasets that mix global and local views, trains one XGBoost classifier per subset, learns fixed ensemble weights by SGD on cross-entropy loss, and then interpolates those weights with per-test-point likelihoods. The result is a method that, in the authors' experiments, keeps high recall on the minority class while retaining decent majority recall, and performs especially well on highly imbalanced data with complex decision boundaries. The paper also claims that the AER framework reduces the asymptotic training complexity of super-linear classifiers and strictly reduces memory complexity for any polynomial-time classifier.

Load-bearing premise

The theoretical speed-up rests on assuming the ensemble weight training stops after very few iterations and on a step in the proof that may be algebraically wrong; if either gives way, the claimed complexity advantage is not established.

Editorial extensions

If this is right

  • Dynamic ensemble methods can be regularized by design rather than by explicit norm penalties, opening a direction for overfitting control in classifier selection.
  • For classifier families whose training time grows super-linearly in the number of instances, wrapping them in AER is claimed to reduce asymptotic training cost, with the benefit growing with the order of m.
  • AER is most advantageous on data with high imbalance ratios and complex minority-class boundaries; on mildly imbalanced benchmarks its gains shrink and some recent baselines remain competitive.
  • The exponential-likelihood variant concentrates weights more sharply, so it tends to do better when training/validation and test distributions are similar, while the log-likelihood variant is safer when the test distribution is less predictable.
  • Regularization from interpolating learned and likelihood-based weights yields better balanced accuracy than either fixed weights or pure dynamic weighting alone.

Reading between the lines

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

  • Editorial inference: if the SGD minimum-norm effect is the active regularizer, then swapping SGD for an explicitly norm-constrained optimizer should produce comparable accuracy, and a controlled comparison would isolate the mechanism.
  • Editorial inference: on minority classes with multiple separated modes, per-Gaussian classifiers should matter more, so AER's advantage over vanilla XGBoost should grow with the number of minority modes; this is testable with GMM-generated data.
  • Editorial inference: the asymptotic memory bound is the more robust theoretical claim because it does not depend on the SGD iteration bound; even if the time-complexity theorem fails, the strict memory advantage may survive.
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

5 major / 5 minor

Summary. The paper proposes AER, an adaptive ensemble selection method for binary imbalanced classification. The method fits a GMM to the majority class, generates 2L subsets (global-geometry and local-geometry balanced subsets), trains XGBoost classifiers on each, learns static ensemble weights via SGD, and interpolates these with likelihood-based dynamic weights at test time. The authors claim that the method provides implicit regularization, reduces asymptotic time and memory complexity relative to vanilla implementations, and outperforms existing algorithms on seven UCI datasets and five GMM-generated datasets, with McNemar and Wilcoxon tests verifying significance. The central evidence is a complexity analysis in Section 4 and single-split experimental comparisons in Sections 5-6.

Significance. If the claims were fully supported, the paper would offer a practically relevant contribution: a regularization mechanism for dynamic ensemble selection that is both cheap and effective on high-imbalance problems. The algorithmic idea is clearly laid out, and the choice of modern baselines such as LightGBM and focal-loss XGBoost is reasonable. However, the manuscript's own evidence does not support the headline claims. The complexity improvement is derived under an unjustified iteration bound and contains an algebraic error, and the experimental tables contain internally inconsistent values; the statistical tests are also mislabeled and report erroneous p-values. The paper would need substantial rework of both the theory and the experiments before its contribution can be assessed.

major comments (5)
  1. [§4.1, Theorem 1] The proof of Theorem 1 contains an algebraic error. In the stochastic-gradient-descent item, the text writes O(mnLt2) = O(mLt2/R), but multiplication by n cannot be replaced by division by R. Substituting the stated assumption t2 ≤ nR/L into O(mnLt2) gives O(m n² R), not the claimed O(mn), so the simplification of TAER(m,n) to O(m^a / L^{a-1} n^b) is not established. The assumption t2 ≤ nR/L is also not justified anywhere in the paper; t2 is the number of SGD iterations, and no bound on it is derived.
  2. [§6.3, Table 8] The Dynamic Ensemble rows in Table 8 are internally inconsistent and duplicate values from Table 4. The (log) row has the same minority recall (83.33%), majority recall (0.47%), and TP-FP ratio (0.8373) as the Plain XGBoost row in Table 4, and the (exp) row matches the Focal-loss XGBoost row on the same three metrics; a GMM-generated dataset would not be expected to reproduce these values. Within Table 8, the two Dynamic Ensemble rows share identical F1 (0.0132) and G-mean (0.0195) despite different majority recalls (0.47% vs 78.19%), contradicting the formulas in Eq. (23); the reported balanced accuracy 48.35% also does not equal the average of the listed recalls. These inconsistencies invalidate the experimental evidence for the GMM dataset and the abstract's claim of widespread outperformance.
  3. [§5.4 and Tables 6, 11] The statistical tests are reported in a way that is not reliable. The Wilcoxon signed-rank test statistic is labeled 'χ2 statistics' in Tables 7, 12, 15, 18, 21, and 24, which is incorrect. More concretely, Table 6 reports p=0.02 for a McNemar χ2 of 54.0, whereas the one-degree-of-freedom chi-square p-value is many orders of magnitude smaller, and Table 11 reports p<0.01 for χ2=4.17, whose correct p-value is approximately 0.04. These errors mean the stated conclusion that 'two hypothesis tests verify the statistical significance' is not supported.
  4. [§5.1–§5.2, §6] All performance numbers are single-split point estimates. The datasets are split once into train/validation/test sets at a 3:1:1 ratio, and no standard deviations, confidence intervals, or repeated runs are reported. For the Bioassay dataset, the cost-sensitive SVM and decision-tree results are taken from [59] with an unknown split and unknown parameters, as the paper states in Section 5.2. Given the small test sets (e.g., Ecoli has 336 samples total), these comparisons cannot support the broad claim that AER-XGBoost outperforms existing algorithms 'in most cases'.
  5. [§7 vs §4] The paper's own discussion undercuts the theoretical complexity claim for the implemented method. Section 7 states that XGBoost's time complexity is not in the polynomial-time regime assumed by Theorem 1, that the AER-XGBoost running time is 'usually longer than expected,' and that the experiments use full covariance matrices instead of the diagonal approximation on which the Section 4 analysis relies. Consequently, even if the algebraic issue in Theorem 1 were repaired, the complexity advantage would apply only to polynomial-time base classifiers with diagonal-covariance GMMs, not to the AER-XGBoost system actually evaluated.
minor comments (5)
  1. [§1] The cross-reference 'Experimental Framework and and results analysis are demonstrated in Section??' is broken and contains a duplicated 'and'.
  2. [Table 25] The caption misspells 'Parameters' as 'Parmaeters'.
  3. [Equation (14)] S(x) is described as a clipping function that floors at 0 and ceilings at 1, but the formula given is a min-max normalization, not a clip; the definition should match the use in Equations (13) and Algorithm 2.
  4. [Tables 4, 5, 8] Several tables use '–' for undefined metrics without a consistent legend, making it unclear whether the value is 0, undefined, or omitted.
  5. [Equation (23)] The F1 and G-mean definitions are stated for a single precision/recall pair, but the text earlier defines class-specific precisions and recalls; the paper should state explicitly whose precision and recall are used in the reported F1 and G-mean values.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the AER predictions are evaluated on held-out test data, and the complexity claim is conditional rather than self-referential.

full rationale

The central derivation chain of the paper is empirical: AER constructs GMM-based subsets, fits XGBoost base classifiers, trains ensemble weights with SGD, and interpolates those weights with test-instance likelihoods (Equations 10-19). The hyperparameters lambda and delta are chosen on a validation partition (Section 5.2, Algorithm 3), and the reported metrics are computed on held-out test partitions; they are not recoverable from the training fit by construction, so the headline performance claim is not circular. The statistical tests compare label predictions on the same test folds, which is an independent check rather than a restatement of the fitted values. The theoretical Section 4 is a conditional asymptotic analysis: Theorem 1 assumes t2 <= n*R/L and m = Omega(mv L) and then hides the SGD term; this makes the theorem dependent on assumptions and the proof contains an algebraic error (O(mnLt2) = O(mLt2/R), confusing the feature dimension n with the imbalance ratio R), but the conclusion is not defined as the assumption, so unsoundness here is a correctness issue, not circularity. Self-citations such as the authors' prior Imbalance-XGBoost [17] are used as a comparison baseline and as background, not as the justification for AER's predictions, so they are not load-bearing. The manuscript's own limitations, including the observation that running time is 'usually longer than expected' and that XGBoost is outside the polynomial-time regime, weaken the complexity claim but do not make it circular. The duplicated rows between Table 8 and Table 4 are internal inconsistencies in the reported empirical record, not a derivation that equals its input. Overall, no step in the claimed derivation reduces to its own inputs by construction.

Assumptions & free parameters 8 free parameters · 6 assumptions · 0 invented entities

The central claim relies on the GMM subset generation, the SGD weight learning, and the test-time likelihood interpolation. The assumptions above are domain assumptions about the data geometry and the transfer of SGD regularization; the free parameters are model selection and tuning choices. No invented entities are introduced.

free parameters (8)
  • Number of GMM components L = Per dataset via BIC (e.g., 8 for Bioassay, 12 for Abalone)
    Determines the number of classifiers 2L; selected by minimum BIC over a hand-specified candidate list, so it is tuned to the data.
  • AIC/BIC mixing weight lambda_A = 0.6
    Fixed by 'experiments demonstrate' (Section 3.3, Eq. 16) without a sensitivity study.
  • Ensemble interpolation weight lambda = Grid-searched on validation per dataset (e.g., 0.25 log, 0.15 exp for Bioassay)
    Bias-variance trade-off between static and dynamic weights; chosen to maximize validation balanced accuracy.
  • Decision threshold delta = Grid-searched on validation (e.g., 0.35 log, 0.30 exp for Bioassay)
    Converts the weighted score into a hard label; tuned on validation.
  • SGD learning rate gamma_t = Not specified precisely
    Recommended to be less than 1/||gradient|| but no schedule or value is reported, so the weight learning is not fully specified.
  • SGD stopping tolerance and Max-Step = Not reported
    Stopping criterion is relative cross-entropy change but the threshold is not given; the t2 bound in Theorem 1 depends on this count.
  • XGBoost hyperparameters (eta, max_depth, num_round) = Tuned per dataset from ranges in Appendix Table 25
    Base classifier capacity affects the AER performance; standard hyperparameter tuning.
  • GMM covariance type = Full covariance in experiments
    Experiments use full covariance, while the complexity analysis assumes diagonal covariance (Section 7 acknowledges the mismatch).
assumptions (6)
  • domain assumption SGD converges to a minimum-l2-norm solution and therefore regularizes the learned ensemble weights (Section 2 and 3.3, citing [47] and [12]).
    The cited results are for unconstrained linear models; the AER uses projected, re-normalized updates on the simplex (Eq. 13), so the transfer is not proved.
  • domain assumption A Gaussian mixture fitted to the majority class captures the global data geometry well enough that likelihood-ranked subsets preserve useful structure (Algorithm 1).
    The regularization claim relies on this geometric faithfulness; no validation of GMM fit quality is provided.
  • domain assumption The likelihood of a test point under the same GMM is a valid competence measure for selecting classifiers (Eq. 18).
    This is the dynamic part of the method; the paper does not compare against other competence measures.
  • domain assumption The base classifiers return probabilistic scores in [0,1] so the cross-entropy loss in Eq. 11 is well-defined.
    Section 3.3 says classifiers give 'class prediction (0 or 1)', which would make the log loss undefined; the method implicitly requires probabilities.
  • ad hoc to paper The number of SGD iterations satisfies t2 <= nR/L (Theorem 1, Section 4.1).
    This bound is introduced solely to make the time complexity term vanish; the algorithm does not guarantee it.
  • standard math The GMM EM algorithm converges in t1 = O(log m) iterations (Section 4.1).
    Super-linear convergence near a local optimum is cited, but the O(log m) bound is an assumption.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Adaptive Ensemble of Classifiers with Regularization for Imbalanced Data Classification." pith.science (2026). https://pith.science/paper/YH7CKW22

@misc{pith2026190803595,
  author       = {Pith},
  title        = {Pith review of: Adaptive Ensemble of Classifiers with Regularization for Imbalanced Data Classification},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YH7CKW22}},
  note         = {Machine review of arXiv:1908.03595}
}
read the original abstract

The dynamic ensemble selection of classifiers is an effective approach for processing label-imbalanced data classifications. However, such a technique is prone to overfitting, owing to the lack of regularization methods and the dependence of the aforementioned technique on local geometry. In this study, focusing on binary imbalanced data classification, a novel dynamic ensemble method, namely adaptive ensemble of classifiers with regularization (AER), is proposed, to overcome the stated limitations. The method solves the overfitting problem through implicit regularization. Specifically, it leverages the properties of stochastic gradient descent to obtain the solution with the minimum norm, thereby achieving regularization; furthermore, it interpolates the ensemble weights by exploiting the global geometry of data to further prevent overfitting. According to our theoretical proofs, the seemingly complicated AER paradigm, in addition to its regularization capabilities, can actually reduce the asymptotic time and memory complexities of several other algorithms. We evaluate the proposed AER method on seven benchmark imbalanced datasets from the UCI machine learning repository and one artificially generated GMM-based dataset with five variations. The results show that the proposed algorithm outperforms the major existing algorithms based on multiple metrics in most cases, and two hypothesis tests (McNemar's and Wilcoxon tests) verify the statistical significance further. In addition, the proposed method has other preferred properties such as special advantages in dealing with highly imbalanced data, and it pioneers the research on the regularization for dynamic ensemble methods.

Figures

Figures reproduced from arXiv: 1908.03595 by the authors.

Figure 1
Figure 1. The overall process of the proposed AER algorithm [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. The trained weights of the algorithm for UCI Bioassay data (round to 2 decimals). The [PITH_FULL_IMAGE:figures/full_fig_p015_2.png] view at source ↗
Figure 3
Figure 3. The performance of the overall algorithm with respect to the change of the interpolation [PITH_FULL_IMAGE:figures/full_fig_p015_3.png] view at source ↗
Figures from the paper (9 more)
Figure 4
Figure 4. Figure 4: The performances of the overall algorithm with respect to the change of the interpolation [PITH_FULL_IMAGE:figures/full_fig_p015_4.png]
Figure 5
Figure 5. Figure 5: The performances of the proposed dynamic ensemble algorithm with respect to the change [PITH_FULL_IMAGE:figures/full_fig_p016_5.png]
Figure 6
Figure 6. Figure 6: The performances of the proposed dynamic ensemble algorithm with respect to the [PITH_FULL_IMAGE:figures/full_fig_p016_6.png]
Figure 7
Figure 7. Figure 7: Change in F1 score and G-mean of the minority data with respect to the changing value of δ in Equation 20 on the UCI Bioassay dataset. Log likelihood is adopted to compute the likelihood of each test data [PITH_FULL_IMAGE:figures/full_fig_p017_7.png]
Figure 8
Figure 8. Figure 8: Change in F1 score and G-mean of the minority data with respect to the changing value of δ in Equation 20 on the UCI Bioassay dataset. Exponential likelihood is adopted to compute the likelihood of each test data. likelihood counterpart yields a better TP-FP ratio, bec…
Figure 9
Figure 9. Figure 9: Performance of the proposed dynamic ensemble algorithm with respect to changes in the [PITH_FULL_IMAGE:figures/full_fig_p017_9.png]
Figure 10
Figure 10. Figure 10: Performances of the proposed dynamic ensemble algorithm with respect to changes in [PITH_FULL_IMAGE:figures/full_fig_p018_10.png]
Figure 11
Figure 11. Figure 11: Performance of the proposed dynamic ensemble algorithm with respect to changes in [PITH_FULL_IMAGE:figures/full_fig_p018_11.png]
Figure 12
Figure 12. Figure 12: Performances of the proposed dynamic ensemble algorithm with respect to changes [PITH_FULL_IMAGE:figures/full_fig_p019_12.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

68 extracted references · 61 canonical work pages

  1. [59]

    Virtual screening of bioassay data.Journal of cheminformatics, 1(1):21, 2009

    Amanda C Schierz. Virtual screening of bioassay data.Journal of cheminformatics, 1(1):21, 2009

    work page 2009

  2. [1]

    Learning from class-imbalanced data: Review of methods and applications.Expert Systems with Applications, 73:220–239, 2017

    Guo Haixiang, Li Yijing, Jennifer Shang, Gu Mingyun, Huang Yuanyue, and Gong Bing. Learning from class-imbalanced data: Review of methods and applications.Expert Systems with Applications, 73:220–239, 2017

    work page 2017

  3. [2]

    A hybrid feature selection with ensemble classification for imbalanced healthcare data: A case study for brain tumor diagnosis.IEEE Access, 4:9145– 9154, 2016

    Shamsul Huda, John Yearwood, Herbert F Jelinek, Mohammad Mehedi Hassan, Giancarlo Fortino, and Michael Buckland. A hybrid feature selection with ensemble classification for imbalanced healthcare data: A case study for brain tumor diagnosis.IEEE Access, 4:9145– 9154, 2016

    work page 2016

  4. [3]

    Atefeh Dehghani Ashkezari, Hui Ma, Tapan K Saha, and Chandima Ekanayake. Application of fuzzy support vector machine for determining the health index of the insulation system of in-service power transformers.IEEE Transactions on Dielectrics and Electrical Insulation, 20(3):965–973, 2013. 31

    work page 2013

  5. [4]

    Rosefw-rf: the winner algorithm for the ecbdl’14 big data competition: an extremely imbalanced big data bioinformatics problem.Knowledge-Based Systems, 87:69–79, 2015

    Isaac Triguero, Sara del Río, Victoria López, Jaume Bacardit, José M Benítez, and Francisco Herrera. Rosefw-rf: the winner algorithm for the ecbdl’14 big data competition: an extremely imbalanced big data bioinformatics problem.Knowledge-Based Systems, 87:69–79, 2015

    work page 2015

  6. [5]

    Rafael M. O. Cruz, Robert Sabourin, and George D. C. Cavalcanti. On dynamic ensemble selection and data preprocessing for multi-class imbalance learning.International Journal of Pattern Recognition and Artificial Intelligence, 33:1940009:1–1940009:29, 2019

    work page 2019

  7. [6]

    A study on combining dynamic selection and data preprocessing for imbalance learning.Neurocomputing, 286:179–192, 2018

    Anandarup Roy, Rafael MO Cruz, Robert Sabourin, and George DC Cavalcanti. A study on combining dynamic selection and data preprocessing for imbalance learning.Neurocomputing, 286:179–192, 2018

    work page 2018

  8. [7]

    Combining multiple algorithms in classifier ensembles using generalized mixture functions.Neurocomputing, 313:402–414, 2018

    Valdigleis S Costa, Antonio Diego S Farias, Benjamin Bedregal, Regivan HN Santiago, and Anne Magaly de P Canuto. Combining multiple algorithms in classifier ensembles using generalized mixture functions.Neurocomputing, 313:402–414, 2018

    work page 2018

Show all 68 references
  1. [8]

    A framework for dynamic classifier selection oriented by the classification problem difficulty

    André L Brun, Alceu S Britto Jr, Luiz S Oliveira, Fabricio Enembreck, and Robert Sabourin. A framework for dynamic classifier selection oriented by the classification problem difficulty. Pattern Recognition, 76:175–190, 2018

    2018

  2. [9]

    Bayesian reasoning and machine learning

    David Barber. Bayesian reasoning and machine learning. Cambridge University Press, 2012

    2012

  3. [10]

    Robust text-independent speaker identification using gaussian mixture speaker models.IEEE transactions on speech and audio processing, 3(1):72–83, 1995

    Douglas A Reynolds and Richard C Rose. Robust text-independent speaker identification using gaussian mixture speaker models.IEEE transactions on speech and audio processing, 3(1):72–83, 1995

    1995

  4. [11]

    Learning characteristics of stochastic-gradient-descent algorithms: A general study, analysis, and critique.Signal processing, 6(2):113–133, 1984

    William A Gardner. Learning characteristics of stochastic-gradient-descent algorithms: A general study, analysis, and critique.Signal processing, 6(2):113–133, 1984

    1984

  5. [12]

    Under- standing deep learning requires rethinking generalization

    Chiyuan Zhang, Samy Bengio, Moritz Hardt, Benjamin Recht, and Oriol Vinyals. Under- standing deep learning requires rethinking generalization. In5th International Conference on Learning Representations (ICLR), 2017

    2017

  6. [13]

    Deep learning in neural networks: An overview

    Jürgen Schmidhuber. Deep learning in neural networks: An overview. Neural networks, 61:85–117, 2015

    2015

  7. [14]

    Optimization methods for large-scale ma- chine learning

    Léon Bottou, Frank E Curtis, and Jorge Nocedal. Optimization methods for large-scale ma- chine learning. Siam Review, 60(2):223–311, 2018

    2018

  8. [15]

    Xgboost: A scalable tree boosting system

    Tianqi Chen and Carlos Guestrin. Xgboost: A scalable tree boosting system. InProceedings of the 22nd ACM SIGKDD international conference on knowledge discovery and data mining, pages 785–794, 2016

    2016

  9. [16]

    Focal loss for dense object detection

    Tsung-Yi Lin, Priyal Goyal, Ross Girshick, Kaiming He, and Piotr Dollár. Focal loss for dense object detection. IEEE transactions on pattern analysis and machine intelligence, 2018

    2018

  10. [17]

    Imbalance-xgboost: Leveraging weighted and focal losses for binary label-imbalanced classification with xgboost.Pattern Recognition Letters, 2020

    Chen Wang, Chengyuan Deng, and Suzhen Wang. Imbalance-xgboost: Leveraging weighted and focal losses for binary label-imbalanced classification with xgboost.Pattern Recognition Letters, 2020

    2020

  11. [18]

    Lightgbm: A highly efficient gradient boosting decision tree

    Guolin Ke, Qi Meng, Thomas Finley, Taifeng Wang, Wei Chen, Weidong Ma, Qiwei Ye, and Tie-Yan Liu. Lightgbm: A highly efficient gradient boosting decision tree. InAdvances in neural information processing systems, pages 3146–3154, 2017

    2017

  12. [19]

    Handling data irregularities in classification: Foundations, trends, and future challenges.Pattern Recognition, 81:674–693, 2018

    Swagatam Das, Shounak Datta, and Bidyut B Chaudhuri. Handling data irregularities in classification: Foundations, trends, and future challenges.Pattern Recognition, 81:674–693, 2018

    2018

  13. [20]

    Analysing the classification of imbalanced data-sets with multiple classes: Binarization techniques and ad-hoc approaches.Knowledge-based systems, 42:97–110, 2013

    Alberto Fernández, Victoria López, Mikel Galar, María José Del Jesus, and Francisco Her- rera. Analysing the classification of imbalanced data-sets with multiple classes: Binarization techniques and ad-hoc approaches.Knowledge-based systems, 42:97–110, 2013

    2013

  14. [21]

    Learning from imbalanced data: open challenges and future directions

    Bartosz Krawczyk. Learning from imbalanced data: open challenges and future directions. Progress in Artificial Intelligence, 5(4):221–232, 2016. 32

    2016

  15. [22]

    Survey of resampling techniques for improving classification performance in unbalanced datasets

    Ajinkya More. Survey of resampling techniques for improving classification performance in unbalanced datasets. arXiv preprint arXiv:1608.06048, 2016

    2016 arXiv

  16. [23]

    Cost-sensitive learning of deep feature representations from imbalanced data.IEEE transactions on neural networks and learning systems, 2017

    Salman H Khan, Munawar Hayat, Mohammed Bennamoun, Ferdous A Sohel, and Roberto Togneri. Cost-sensitive learning of deep feature representations from imbalanced data.IEEE transactions on neural networks and learning systems, 2017

    2017

  17. [24]

    Near-bayesian support vector machines for imbalanced data classification with equal or unequal misclassification costs.Neural Networks, 70:39–52, 2015

    Shounak Datta and Swagatam Das. Near-bayesian support vector machines for imbalanced data classification with equal or unequal misclassification costs.Neural Networks, 70:39–52, 2015

    2015

  18. [25]

    Class-specific extreme learning machine for handling binary class imbalance problem.Neural Networks, 105:206–217, 2018

    Bhagat Singh Raghuwanshi and Sanyam Shukla. Class-specific extreme learning machine for handling binary class imbalance problem.Neural Networks, 105:206–217, 2018

    2018

  19. [26]

    Online sequential class-specific extreme learning machine for binary imbalanced learning.Neural Networks, 119:235–248, 2019

    Sanyam Shukla and Bhagat Singh Raghuwanshi. Online sequential class-specific extreme learning machine for binary imbalanced learning.Neural Networks, 119:235–248, 2019

    2019

  20. [27]

    One-class versus binary classi- fication: Which and when? In Machine Learning and Applications (ICMLA), 2012 11th International Conference on, volume 2, pages 102–106

    Colin Bellinger, Shiven Sharma, and Nathalie Japkowicz. One-class versus binary classi- fication: Which and when? In Machine Learning and Applications (ICMLA), 2012 11th International Conference on, volume 2, pages 102–106. IEEE, 2012

    2012

  21. [28]

    Feature learning with a divergence-encouraging autoencoder for imbalanced data classification.IEEE Access, 6:70197–70211, 2018

    Ruisen Luo, Qian Feng, Chen Wang, Xiaomei Yang, Haiyan Tu, Qin Yu, Shaomin Fei, and Xiaofeng Gong. Feature learning with a divergence-encouraging autoencoder for imbalanced data classification.IEEE Access, 6:70197–70211, 2018

    2018

  22. [29]

    Bagging and boosting.Encyclopedia of Biostatistics, 1, 2005

    Richard De Veaux. Bagging and boosting.Encyclopedia of Biostatistics, 1, 2005

    2005

  23. [30]

    An overview of ensemble methods for binary classifiers in multi-class problems: Experimental study on one-vs-one and one-vs-all schemes.Pattern Recognition, 44(8):1761– 1776, 2011

    Mikel Galar, Alberto Fernández, Edurne Barrenechea, Humberto Bustince, and Francisco Herrera. An overview of ensemble methods for binary classifiers in multi-class problems: Experimental study on one-vs-one and one-vs-all schemes.Pattern Recognition, 44(8):1761– 1776, 2011

    2011

  24. [31]

    An empirical study of learning from imbalanced data using random forest

    Taghi M Khoshgoftaar, Moiz Golawala, and Jason Van Hulse. An empirical study of learning from imbalanced data using random forest. In19th IEEE international conference on Tools with Artificial Intelligence, ICTAI., volume 2, pages 310–317. IEEE, 2007

    2007

  25. [32]

    A novel ensemble method for imbalanced data learning: bagging of extrapolation-smote svm.Computational intelligence and neuroscience, pages 1–11, 2017

    Qi Wang, ZhiHao Luo, JinCai Huang, YangHe Feng, and Zhong Liu. A novel ensemble method for imbalanced data learning: bagging of extrapolation-smote svm.Computational intelligence and neuroscience, pages 1–11, 2017

    2017

  26. [33]

    A review on ensembles for the class imbalance problem: bagging-, boosting-, and hybrid-based approaches

    Mikel Galar, Alberto Fernandez, Edurne Barrenechea, Humberto Bustince, and Francisco Herrera. A review on ensembles for the class imbalance problem: bagging-, boosting-, and hybrid-based approaches. IEEE Transactions on Systems, Man, and Cybernetics, Part C (Applications and R...

    2012

  27. [34]

    A survey of multiple classifier systems as hybrid systems.Information Fusion, 16:3–17, 2014

    Michał Woźniak, Manuel Graña, and Emilio Corchado. A survey of multiple classifier systems as hybrid systems.Information Fusion, 16:3–17, 2014

    2014

  28. [35]

    Comparing boosting and bagging techniques with noisy and imbalanced data.IEEE Transactions on Systems, Man, and Cybernetics-Part A: Systems and Humans, 41(3):552–568, 2011

    Taghi M Khoshgoftaar, Jason Van Hulse, and Amri Napolitano. Comparing boosting and bagging techniques with noisy and imbalanced data.IEEE Transactions on Systems, Man, and Cybernetics-Part A: Systems and Humans, 41(3):552–568, 2011

    2011

  29. [36]

    Improving classifiers and regions of competence in dynamic ensemble selection

    Tiago Pessoa Ferreira De Lima, Anderson Tenorio Sergio, and Teresa Bernarda Ludermir. Improving classifiers and regions of competence in dynamic ensemble selection. InBrazilian Conference on Intelligent Systems, 2014

    2014

  30. [37]

    Philip Kegelmeyer, and Kevin Bowyer

    Kevin Woods, W. Philip Kegelmeyer, and Kevin Bowyer. Combination of multiple classifiers using local accuracy estimates. IEEE transactions on pattern analysis and machine intelli- gence, 19(4):405–410, 1997

    1997

  31. [38]

    From dynamic classifier selection to dynamic ensemble selection.Pattern Recognition, 41(5):1718–1731, 2008

    Albert HR Ko, Robert Sabourin, and Alceu Souza Britto Jr. From dynamic classifier selection to dynamic ensemble selection.Pattern Recognition, 41(5):1718–1731, 2008. 33

    2008

  32. [39]

    Libd3c: ensemble classifiers with a clustering and dynamic selection strategy

    Chen Lin, Wenqiang Chen, Cheng Qiu, Yunfeng Wu, Sridhar Krishnan, and Quan Zou. Libd3c: ensemble classifiers with a clustering and dynamic selection strategy. Neurocom- puting, 123:424–435, 2014

    2014

  33. [40]

    Meta-des: a dynamic ensemble selection framework using meta-learning.Pattern recognition, 48(5):1925– 1935, 2015

    Rafael MO Cruz, Robert Sabourin, George DC Cavalcanti, and Tsang Ing Ren. Meta-des: a dynamic ensemble selection framework using meta-learning.Pattern recognition, 48(5):1925– 1935, 2015

    1925

  34. [41]

    Dynamic classifier ensemble model for customer classification with imbalanced class distribution.Expert Systems with Applications, 39(3):3668–3675, 2012

    Jin Xiao, Ling Xie, Changzheng He, and Xiaoyi Jiang. Dynamic classifier ensemble model for customer classification with imbalanced class distribution.Expert Systems with Applications, 39(3):3668–3675, 2012

    2012

  35. [42]

    Dynamic ensemble selection for multi-class classification with one-class classifiers.Pattern Recognition, 83:34–51, 2018

    Bartosz Krawczyk, Mikel Galar, Michał Woźniak, Humberto Bustince, and Francisco Herrera. Dynamic ensemble selection for multi-class classification with one-class classifiers.Pattern Recognition, 83:34–51, 2018

    2018

  36. [43]

    Noiseout: A simple way to prune neural networks.arXiv preprint arXiv:1611.06211, 2016

    Mohammad Babaeizadeh, Paris Smaragdis, and Roy H Campbell. Noiseout: A simple way to prune neural networks.arXiv preprint arXiv:1611.06211, 2016

    2016 arXiv

  37. [44]

    Dropout: a simple way to prevent neural networks from overfitting.The Journal of Machine Learning Research, 15(1):1929–1958, 2014

    Nitish Srivastava, Geoffrey Hinton, Alex Krizhevsky, Ilya Sutskever, and Ruslan Salakhut- dinov. Dropout: a simple way to prevent neural networks from overfitting.The Journal of Machine Learning Research, 15(1):1929–1958, 2014

    1929

  38. [45]

    Exploratory undersampling for class-imbalance learning

    Xu-Ying Liu, Jianxin Wu, and Zhi-Hua Zhou. Exploratory undersampling for class-imbalance learning. IEEE Transactions on Systems, Man, and Cybernetics, Part B (Cybernetics), 39(2):539–550, 2008

    2008

  39. [46]

    An overlap-sensitive margin classifier for imbalanced and overlapping data.Expert Systems with Applications, 98:72–83, 2018

    Han Kyu Lee and Seoung Bum Kim. An overlap-sensitive margin classifier for imbalanced and overlapping data.Expert Systems with Applications, 98:72–83, 2018

    2018

  40. [47]

    Iterative regularization for learning with convex loss functions

    Junhong Lin, Lorenzo Rosasco, and Ding-Xuan Zhou. Iterative regularization for learning with convex loss functions. The Journal of Machine Learning Research, 17(1):2718–2755, 2016

    2016

  41. [48]

    Generalization properties and im- plicit regularization for multiple passes sgm

    Junhong Lin, Raffaello Camoriano, and Lorenzo Rosasco. Generalization properties and im- plicit regularization for multiple passes sgm. In International Conference on International Conference on Machine Learning, 2016

    2016

  42. [49]

    Evolutionary under- sampling boosting for imbalanced classification of breast cancer malignancy.Applied Soft Computing, 38:714–726, 2016

    Bartosz Krawczyk, Mikel Galar, Łukasz Jeleń, and Francisco Herrera. Evolutionary under- sampling boosting for imbalanced classification of breast cancer malignancy.Applied Soft Computing, 38:714–726, 2016

    2016

  43. [50]

    Elblocker: Predicting blocking bugs with ensemble imbalance learning.Information and Software Technology, 61:93– 106, 2015

    Xin Xia, David Lo, Emad Shihab, Xinyu Wang, and Xiaohu Yang. Elblocker: Predicting blocking bugs with ensemble imbalance learning.Information and Software Technology, 61:93– 106, 2015

    2015

  44. [51]

    Machine learning based mobile malware detection using highly imbalanced network traffic

    Zhenxiang Chen, Qiben Yan, Hongbo Han, Shanshan Wang, Lizhi Peng, Lin Wang, and Bo Yang. Machine learning based mobile malware detection using highly imbalanced network traffic. Information Sciences, 433:346–364, 2018

    2018

  45. [52]

    A new method for occupational fraud detection in process aware information systems

    Shahla Mardani and Hamid Reza Shahriari. A new method for occupational fraud detection in process aware information systems. In10th International ISC Conference on Information Security and Cryptology (ISCISC), pages 1–5. IEEE, 2013

    2013

  46. [53]

    On convergence properties of the em algorithm for gaussian mixtures

    Lei Xu and Michael I Jordan. On convergence properties of the em algorithm for gaussian mixtures. Neural computation, 8(1):129–151, 1996

    1996

  47. [54]

    Scikit-learn: Machine learning in python.Journal of machine learning research, 12(Oct):2825– 2830, 2011

    Fabian Pedregosa, Gaël Varoquaux, Alexandre Gramfort, Vincent Michel, Bertrand Thirion, Olivier Grisel, Mathieu Blondel, Peter Prettenhofer, Ron Weiss, Vincent Dubourg, et al. Scikit-learn: Machine learning in python.Journal of machine learning research, 12(Oct):2825– 2830, 2011

    2011

  48. [55]

    Sensitivity and specificity of information criteria

    John J Dziak, Donna L Coffman, Stephanie T Lanza, and Runze Li. Sensitivity and specificity of information criteria. The Methodology Center and Department of Statistics, Penn State, The Pennsylvania State University, 16(30):140, 2012. 34

    2012

  49. [56]

    An experiment with the edited nearest-neighbor rule.IEEE Transactions on Systems, Man, and Cybernetics, 1976

    Ivan Tomek. An experiment with the edited nearest-neighbor rule.IEEE Transactions on Systems, Man, and Cybernetics, 1976

    1976

  50. [57]

    Friedman

    Jerome H. Friedman. Greedy function approximation: A gradient boosting machine.Annals of Statistics, 29(5):1189–1232, 2001

    2001

  51. [58]

    Keel 3.0: an open source software for multi-stage analysis in data mining

    Isaac Triguero, Sergio González, Jose M Moyano, Salvador García López, Jesús Alcalá Fer- nández, Julián Luengo Martín, Alberto Fernández Hilario, María José del Jesús Díaz, Luciano Sánchez, Francisco Herrera Triguero, et al. Keel 3.0: an open source software for multi-stage an...

    2017

  52. [60]

    Espíndola and Nelson F.F

    Rogério P. Espíndola and Nelson F.F. Ebecken. On extending f-measure and g-mean metrics to multi-class problems.WIT Transactions on Information and Communication Technologies, 35, 2005

    2005

  53. [61]

    Kernel-based extreme learning machine for remote-sensing image classification.Remote Sensing Letters, 4(9):853–862, 2013

    Mahesh Pal, Aaron E Maxwell, and Timothy A Warner. Kernel-based extreme learning machine for remote-sensing image classification.Remote Sensing Letters, 4(9):853–862, 2013

    2013

  54. [62]

    Individual comparisons by ranking methods

    Frank Wilcoxon. Individual comparisons by ranking methods. InBreakthroughs in statistics, pages 196–202. Springer, 1992

    1992

  55. [63]

    Approximate statistical tests for comparing supervised classification learning algorithms

    Thomas G Dietterich. Approximate statistical tests for comparing supervised classification learning algorithms. Neural computation, 10(7):1895–1923, 1998

    1923

  56. [64]

    Nonparametric statistical analysis of machine learning algorithms for regression problems

    Magdalena Graczyk, Tadeusz Lasota, Zbigniew Telec, and Bogdan Trawiński. Nonparametric statistical analysis of machine learning algorithms for regression problems. InInternational Conference on Knowledge-Based and Intelligent Information and Engineering Systems, pages 111–120....

    2010

  57. [65]

    Statsmodels: Econometric and statistical modeling with python

    Skipper Seabold and Josef Perktold. Statsmodels: Econometric and statistical modeling with python. In9th Python in Science Conference, 2010

    2010

  58. [66]

    Robust mixture modelling using the t distribution

    David Peel and Geoffrey J McLachlan. Robust mixture modelling using the t distribution. Statistics and computing, 10(4):339–348, 2000

    2000

  59. [67]

    Scalar quantization as sparse least square optimization.IEEE transactions on pattern analysis and machine intelligence, in press, DOI: 10.1109/TPAMI.2019.2952096

    Chen Wang, Xiaomei Yang, Shaomin Fei, Kai Zhou, Xiaofeng Gong, Miao Du, and Ruisen Luo. Scalar quantization as sparse least square optimization.IEEE transactions on pattern analysis and machine intelligence, in press, DOI: 10.1109/TPAMI.2019.2952096

  60. [68]

    Theeffectiveness of lloyd-type methods for the k-means problem.Journal of the ACM (JACM), 59(6):1–22, 2013

    RafailOstrovsky, YuvalRabani, LeonardJSchulman, andChaitanyaSwamy. Theeffectiveness of lloyd-type methods for the k-means problem.Journal of the ACM (JACM), 59(6):1–22, 2013. 35 A Hyper-parameters for Experiments In this section, we provide hyper-parameter sets for experiments ...

    2013

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.