Pith. sign in

REVIEW 5 major objections 7 minor 48 references

AdaptiveMDL-GenClust: A Robust Clustering Framework Integrating Normalized Mutual Information and Evolutionary Algorithms

T0 review · 5 major / 7 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read A genetic algorithm guided by attribute-weighted MDL consistently beats six classical clustering methods on thirteen benchmarks.

desk verdict The paper's own tables contradict its headline claim, and without code or working equations the contribution is not assessable; desk-reject rather than send to referees. read the letter →

arxiv 2412.05305 v3 pith:YRYVGBII submitted 2024-11-26 cs.NE cs.LG

H. Jahani , F. Zamio This is my paper · ORCID
classification cs.NEcs.LG
keywords minimumdescriptionlengthgeneticalgorithmclusteringnormalizedmutualinformationensembleattributeweightingadjustedRandindex
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 clustering method can be made markedly more accurate and stable by optimizing not a similarity measure but the compressed description length of the data, and that a genetic algorithm can carry out that optimization effectively. The proposed Genetic MDL framework generates a starting partition from an ensemble of k-means runs, then refines it with three genetic phases guided by an attribute-weighted MDL criterion. On thirteen benchmark datasets, the paper reports that the method beats k-means, four hierarchical linkage variants, and fuzzy c-means on all four validation metrics, averaged over 100 independent runs. If this holds, clustering would gain a relatively tuning-light, data-adaptive recipe that is less hostage to initialization and feature scaling.

What carries the argument

The carrying object is the Attribute Weighted Description Length (AWDL) criterion, written as $L' = \{S_m, S_d\}$, where $S_m$ is the weighted sum of cluster-mean magnitudes and $S_d$ is the weighted average absolute deviation of samples from their cluster means, with attribute weights derived from variance. The agreement matrix $A = H H'$ records how often pairs of samples are co-clustered by the input ensemble and supplies both the initial solution and the fitness signal for the genetic phases. ABMDLGAO moves samples between clusters according to agreement-based probabilities, EPMDLGAO does the same with uniform probabilities, and EPAFGAO maximizes a thresholded agreement reward.

What would settle it

Re-run the thirteen-dataset comparison with the cluster count estimated by the MDL criterion itself instead of the known true class count; the central claim stands only if the Genetic MDL advantage over k-means and linkage baselines persists when the number of clusters is not supplied.

Watch

Extended reading notes

Core claim

The central claim is that replacing a similarity-only clustering objective with an attribute-weighted description-length objective, optimized by a genetic algorithm, removes much of the bias and initialization dependence of conventional clustering. Concretely, the paper asserts that its Genetic MDL framework, initialized by a consensus partition from repeated k-means runs and refined by three genetic phases named ABMDLGAO, EPMDLGAO, and EPAFGAO, consistently outperforms k-means, single/average/complete/Ward linkage, and fuzzy c-means on thirteen datasets according to accuracy, NMI, Fisher score, and adjusted Rand index, averaged over 100 runs.

Load-bearing premise

The comparison hands every algorithm the true number of classes for each dataset, and the paper says so explicitly in the Data Sets section, so the reported robust and adaptive performance is not tested in the fully unsupervised setting the title advertises.

Editorial extensions

If this is right

  • Clustering objectives can be reformulated as compression problems: minimizing $L'$ should yield partitions that are at once simple and faithful, usable in place of k-means inertia or linkage criteria.
  • Because the framework starts from an ensemble and evolves it, the method should be more stable across random restarts than k-means, whose results vary with initialization.
  • Variance-based attribute weighting gives the method a built-in feature-relevance adjustment, which should help on datasets with noisy or irrelevant attributes.
  • The same pipeline can be retargeted to high-stakes domains named in the paper, such as EEG and ECG signal grouping and spatial transcriptomics, where stable clusters feed downstream diagnosis.

Reading between the lines

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

  • Beyond the paper: since the agreement matrix is built from k-means runs, the framework's view of consensus inherits k-means' biases; swapping in diverse generators such as density- or model-based clustering is a natural test of whether the MDL refinement, rather than the ensemble source, drives the gains.
  • Beyond the paper: all four metrics are measured with the true number of clusters supplied; a fair unsupervised comparison would estimate the cluster count from the data, for example by minimizing $L'$ itself, and that is where the reported margin could shrink or vanish.
  • Beyond the paper: the reported absolute scores on several datasets are low, so 'consistently outperforms' is a relative statement; the method may improve on classical baselines without yet being a universal clusterer on hard data.
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 / 7 minor

Summary. The paper proposes AdaptiveMDL-GenClust, a clustering framework that combines an ensemble-based initial solution with the Minimum Description Length (MDL) principle, attribute-weighted description length (AWDL), and three genetic-algorithm optimization stages (ABMDLGAO, EPMDLGAO, EPAFGAO). The authors claim that this framework consistently outperforms k-means, four hierarchical linkage methods, and fuzzy c-means on thirteen UCI benchmark datasets, with higher accuracy, improved stability, and reduced bias. The manuscript provides algorithmic outlines, mathematical equations for the proposed objective functions, and experimental tables with means and standard deviations over 100 runs, plus a ranking analysis based on a t-test. The evaluation uses four metrics: accuracy, NMI, Fisher score, and ARI.

Significance. If fully supported, a clustering framework that automatically adapts to dataset structure and outperforms standard baselines across diverse benchmarks would be a meaningful contribution to evolutionary and MDL-based clustering. The use of thirteen datasets and four validation metrics, with 100-run averages, is a reasonable evaluation scale, and the idea of combining MDL-style description length with an ensemble initialization is worth investigating. However, the manuscript as written does not establish the claimed result: the headline claim is directly contradicted by the paper's own tables, the method is not reproducible from the provided equations and missing hyperparameters, the experimental protocol relies on true labels and true cluster counts, and the statistical analysis is incomplete. No code or data are provided. The potential significance is therefore not realized in the current manuscript.

major comments (5)
  1. [Experimental Results, Tables 2-5] The central claim that the proposed framework 'consistently outperforms' traditional clustering methods is contradicted by the paper's own numbers. In Table 2, balance_scale NMI is 12.39 for k-means and 11.63 for FCM, while ABMDLGAO, EPMDLGAO, and EPAFGAO score 9.69, 9.28, and 8.26, respectively. On Nbalance_scale, k-means NMI is 11.51, above all three proposed variants. Table 4 shows Nwine accuracy of 96.63 for both k-means and FCM, whereas ABMDLGAO is 88.79 and EPAFGAO is 67.45. The 'improved stability' claim is also unsupported: Table 2 reports standard deviations such as 26.57 ± 40.53 for ABMDLGAO on NYeast and 6.06 ± 21.91 for EPAFGAO on Nbreast. The paper never defines an aggregate 'Genetic MDL' result, so no reasonable reading of the tables supports consistent superiority.
  2. [Data Sets, Table 1] The evaluation protocol states that 'the number of clusters and the true labels of the samples were known beforehand,' and this knowledge is used to set the number of clusters and to compute accuracy, NMI, and ARI. This is a supervised setting, not the unsupervised clustering setting advertised in the abstract. Moreover, the Agreement Matrix (Figure 3) and the initial solution C0 (Figure 4) are built using the same k-means algorithm that later serves as a baseline, so the comparison is not independent of the ensemble construction. The paper provides no experiments for the fully unsupervised case in which k is unknown, despite claiming adaptability and reduced dependency on initial inputs.
  3. [The Proposed Method, Eqs. (2), (9), (12), (13), (16), (20), (22)] Several central equations are malformed or contain undefined symbols, making the method irreproducible. Equation (2) misindexes the normalization formula: X_ij is described both as the jth sample and as the value of the jth attribute of the ith sample. Equation (9) omits the weights w_j on all terms after the first absolute difference. Equation (12) defines probabilities only for the cases max(A_iq) < max(A_i) and max(A_iq) = max(A_i), leaving the case max(A_iq) > max(A_i) undefined, while Eq. (13) defines V_p = 1 − V without defining V. Equation (20) uses inconsistent notation in the NMI definition, and Eq. (22) writes ARI with symbols M, C, and combinatorial terms that are not formally defined. Without corrected equations, the proposed objective functions cannot be implemented or checked.
  4. [Genetic algorithm Multiple Optimization framework and Figure 2] The genetic algorithm itself is not described. The paper lists three fitness functions (ABMDLGAO, EPMDLGAO, EPAFGAO) but provides no details on population initialization, selection, crossover, mutation, elitism, or termination criteria beyond 'converged or the predefined number of iterations NK is reached.' No values are given for NK, ensemble size, random subset fraction e, or any GA parameter. The 'quasi-code' in Figure 2 only says 'Use the Genetic algorithm to start to optimize,' which is not an algorithm specification. Because the central claim depends on this optimization, this omission is load-bearing.
  5. [Algorithm ranking, Tables 6-9] The statistical analysis is incomplete. The text describes a paired t-test, but Tables 6-9 report only integer 'priorities' ranging from -8 to 8, with no test statistic, p-value, or confidence interval, and no aggregate win/loss counts across the 13 datasets. The claimed 'T-test results' therefore do not demonstrate that any difference is statistically significant, and the tables cannot be independently verified from the stated formulas.
minor comments (7)
  1. [Throughout] The paper uses 'quasi-code' where 'pseudo-code' is standard; please correct this in the text and figure captions.
  2. [Figure 1] Figure 1 is mentioned in the text but its diagram is not actually included in the manuscript, so the described framework cannot be visually inspected.
  3. [Table 1] The dataset name 'Nionosphere' likely refers to 'Ionosphere'; please use the standard UCI dataset name.
  4. [Equation (18) and surrounding text] There are spelling inconsistencies: 'F-mesture', 'F − meature', and 'Fisher standard' all appear; please unify the terminology.
  5. [Equation (20)] The NMI formula uses undefined symbols such as p_i1, p_i, and p_j; please restate with standard contingency-table notation and define all terms.
  6. [References] Reference [40] spells the author as 'Graunwald'; the correct spelling is 'Grünwald'.
  7. [Figure 4] The pseudo-code contains a typo, 'Patr1', which should be 'Part1'.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the reported comparisons are empirical and the optimization objectives are not defined in terms of the evaluation metrics.

full rationale

The derivation chain is not circular. The proposed framework defines an AWDL objective L' = {Sm, Sd} (Eqs. 7-9) directly from normalized data and variance-based attribute weights, and the three genetic-optimization variants minimize L' or maximize the agreement function F(C) (Eq. 17). None of these objectives is defined in terms of the reported outcome metrics (accuracy, NMI, Fisher, ARI), which are computed against the known true labels of the UCI benchmark sets; the reported comparisons are therefore empirical, not tautological. The agreement matrix A (Fig. 3) and initial solution C0 are produced by k-means runs, and k-means is also a baseline, which is a legitimate fairness and independence concern for the ensemble comparison, but it does not make the reported NMI, ARI, accuracy, or Fisher values equal to the optimization inputs by construction: EPMDLGAO and ABMDLGAO optimize a description-length criterion, and even EPAFGAO's agreement fitness is a proxy objective, with the reported metrics measured against true labels only during evaluation. There are no load-bearing self-citations (the Parvin et al. references are not by the present authors and are not used to justify the framework's core), no imported uniqueness theorem, and no fitted parameter later relabeled as a prediction. The central claim of consistent superiority is a strong empirical assertion that the tables may not support, but that failure would be one of evidence or correctness, not circularity.

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

The method relies on six hand-chosen or unreported parameters, including the true cluster count, the 0.6 threshold, ensemble size, subset fraction, genetic algorithm settings, and stopping thresholds. The central empirical claim rests on domain assumptions linking the AWDL objective to label accuracy, assuming the ensemble is informative, and assuming variance-based feature weighting is valid. No new physical or conceptual entities are introduced.

free parameters (6)
  • Number of clusters k = True class count per dataset (e.g., 3 for Iris, 10 for NYeast)
    Set by hand from ground-truth labels for all datasets; the method does not estimate k, which is a form of supervision in an advertised unsupervised method.
  • Consensus threshold coefficient = 0.6
    Equation 14 defines B as 0.6 times (max(A) minus min(A)) plus min(A); the 0.6 constant is chosen without sensitivity analysis.
  • Ensemble size = not reported
    Fig. 3 loops over EnsembleSize but the paper never states the value, though the agreement matrix depends on it.
  • Random subset fraction for initial solution = not reported (example: 80%)
    Fig. 4 uses a subset Part1; the text mentions 80% as an example but does not report the actual fraction used.
  • GA population size, crossover and mutation rates = not reported
    These standard genetic algorithm parameters are never given, although they control the optimization outcome.
  • Maximum iterations NK and convergence thresholds = not reported
    The stopping conditions for ABMDLGAO, EPMDLGAO, and EPAFGAO are referenced in Fig. 2 but their numeric values are absent.
assumptions (5)
  • domain assumption Minimizing the AWDL criterion L' (Eqs. 8-9) produces clusterings that agree with true class labels.
    The paper uses L' as the fitness to be minimized by the genetic search, but no derivation or experiment links lower L' to higher accuracy or NMI; the claim that this yields superior performance rests on this unproved correspondence.
  • domain assumption The number of clusters is known and equals the true class count.
    The Data Sets section states the number of clusters and true labels are known beforehand; all experiments set k to this value, so the evaluation does not test the unsupervised setting claimed by the abstract.
  • domain assumption The ensemble of k-means runs used to build the agreement matrix A is sufficiently diverse and informative.
    Fig. 3 builds A only from repeated k-means clustering; if these runs are highly correlated, the consensus matrix will not add independent information, yet the method's later stages depend on A.
  • ad hoc to paper Attribute variance is a valid proxy for attribute informativeness.
    Equations 3-5 assign weights proportional to variance without any evidence that high-variance attributes are more useful for clustering; this weighting scheme is introduced specifically for this method.
  • standard math MDL is an appropriate model-selection criterion for clustering.
    The paper invokes the MDL principle from reference [40] as a general inference framework; this is an accepted background theory, though its specific application here is not derived.

how reviews work

0 comments
Cite this review

Pith. "Pith review of AdaptiveMDL-GenClust: A Robust Clustering Framework Integrating Normalized Mutual Information and Evolutionary Algorithms." pith.science (2026). https://pith.science/paper/YRYVGBII

@misc{pith2026241205305,
  author       = {Pith},
  title        = {Pith review of: AdaptiveMDL-GenClust: A Robust Clustering Framework Integrating Normalized Mutual Information and Evolutionary Algorithms},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YRYVGBII}},
  note         = {Machine review of arXiv:2412.05305}
}
read the original abstract

Clustering algorithms are pivotal in data analysis, enabling the organization of data into meaningful groups. However, individual clustering methods often exhibit inherent limitations and biases, preventing the development of a universal solution applicable to diverse datasets. To address these challenges, we introduce a robust clustering framework that integrates the Minimum Description Length (MDL) principle with a genetic optimization algorithm. The framework begins with an ensemble clustering approach to generate an initial clustering solution, which is then refined using MDL-guided evaluation functions and optimized through a genetic algorithm. This integration allows the method to adapt to the dataset's intrinsic properties, minimizing dependency on the initial clustering input and ensuring a data-driven, robust clustering process. We evaluated the proposed method on thirteen benchmark datasets using four established validation metrics: accuracy, normalized mutual information (NMI), Fisher score, and adjusted Rand index (ARI). Experimental results demonstrate that our approach consistently outperforms traditional clustering methods, yielding higher accuracy, improved stability, and reduced bias. The methods adaptability makes it effective across datasets with diverse characteristics, highlighting its potential as a versatile and reliable tool for complex clustering tasks. By combining the MDL principle with genetic optimization, this study offers a significant advancement in clustering methodology, addressing key limitations and delivering superior performance in varied applications.

Figures

Figures reproduced from arXiv: 2412.05305 by the authors.

Figure 2
Figure 2. [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. shows the Agreement matrix formation algorithm. for i=1:EnsembleSize IDX=kmeans(X,ClusterNum); out=[out,IDX]; end for t=1:size(out,2) for i=1:size(out,2) B=zeros(size(out,1),1); for j=1:size(out,1) if out(j,t)==i B(j,1)=1; end end H=[H,B]; end end A=H*H'; Fig.3. Agreement matrix formation algorithm The formation of the Agreement Matrix (A) can be explained through an example. Consider three individual clustering alg… view at source ↗
Figure 7
Figure 7. Fig.7 [PITH_FULL_IMAGE:figures/full_fig_p017_7.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

48 extracted references · 48 canonical work pages

  1. [1]

    Data clustering: a review,

    Jain, A.K., Murty, M.N., and Flynn, P.J., “Data clustering: a review,” ACM Comput. Surv. Vol. 31, no. 3, pp. 264-323, 1999

    work page 1999

  2. [2]

    Survey of clustering data mining techniques,

    Berkhin, P., “Survey of clustering data mining techniques,” Technical Report, Accrue Software, 2002

    work page 2002

  3. [3]

    Survey of clustering algorithms,

    Xu, R., and Wunsch, D., “Survey of clustering algorithms,” II, IEEE Transactions on Neural Networks, vol. 16, no. 3, pp. 645- 678, 2005

    work page 2005

  4. [4]

    Computational Approaches for Protein Function Prediction,

    Pandey, G., Kumar, V. and Steinbach, M., “Computational Approaches for Protein Function Prediction,” Supported by the National Science Foundation under Grant Nos. IIS-0308264 and ITR-0325949, 2007

    work page 2007

  5. [5]

    Algorithms for Clustering Data,

    Jain, A.K. and Dubes, R.C., “Algorithms for Clustering Data,” Prentice Hall, 1988

    work page 1988

  6. [6]

    Pattern Classification,

    Duda, R., Hart, P. and Stork, D., “Pattern Classification,” John Wiley & Sons, 2001

    work page 2001

  7. [7]

    The Elements of Statistical Learning: Data Mining, Inference and Prediction,

    Hastie, T., Tibshirani, R., and Friedman, J., “The Elements of Statistical Learning: Data Mining, Inference and Prediction,” Springer, 2001

    work page 2001

  8. [8]

    Cluster Analysis for Gene Expression Data: A Survey,

    Jiang, D., Tang, C. and Zhang, A., "Cluster Analysis for Gene Expression Data: A Survey," IEEE Transactions on Knowledge and Data Engineering, vol. 16, no. 11, pp. 1370 -1386, Nov. 2004

    work page 2004

Show all 48 references
  1. [9]

    and Ghosh J

    Strehl A. and Ghosh J. (2002), Cluster ensembles - a knowledge reuse framework for combining multiple partitions. Journal of Machine Learning Research, 3(Dec):583–617

    2002

  2. [10]

    Cluster Analysis,

    Everitt, B., “Cluster Analysis,” Third Edition. Edward Arnold: a member of the Hodder Headline Group. 1993

    1993

  3. [11]

    Hierarchical Clustering Schemes,

    Johnson, S.C., "Hierarchical Clustering Schemes," Psychometrika, vol. 2, pp. 241 -254, 1967

    1967

  4. [12]

    Statistical Pattern Recognition,

    Webb, A., “Statistical Pattern Recognition,” Arnold: a member of the Hodder Headline Group, pp. 275 - 317, 1999

    1999

  5. [13]

    Some methods for classification and analysis of multivariate observations,

    Mcqueen, J., “Some methods for classification and analysis of multivariate observations,” In Proceedings of the Fifth Berkeley Symposium on Mathematical Statistics and Probability, pp. 281 –297, 1967

    1967

  6. [14]

    Finding Groups in Data: An Introduction to Cluster Analysis,

    Kaufman, L. and Rousseeuw, P.J., “Finding Groups in Data: An Introduction to Cluster Analysis,” Wiley - Interscience: New York (Series in Applied Probability and Statistics), ISBN 0 -471-87876-6, 1990

    1990

  7. [15]

    A Density-Based Algorithm for Discovering Clusters in Large Spatial Databases with Noise,

    Ester, M., Kriegel, H.P. and Xu, X., “A Density-Based Algorithm for Discovering Clusters in Large Spatial Databases with Noise,” In: Proc. 2nd ACM SIGKDD, pp. 226-231, 1996

    1996

  8. [16]

    OPTICS: Ordering Points to Identify the Clustering Structure,

    Ankerst, M., Breunig, M., Kriegel, H.P. and Sander, J., “OPTICS: Ordering Points to Identify the Clustering Structure,” Management of Data Mining, pp. 49–60, 1999

    1999

  9. [17]

    STING: A Statistical Information Grid Approach to Spatial Data Mining,

    Wang, W., Yang, J. and Muntz, R., “STING: A Statistical Information Grid Approach to Spatial Data Mining,” In: Proc. of the 23rd Very Large Databases Conf, 1997

    1997

  10. [18]

    Parvin, H., Parvin, S. (2013). A classifier ensemble of binary classifier ensembles. International Journal of Learning Management Systems, 1(2), 36-46

    2013

  11. [19]

    Automatic formation of topological maps of patterns in a self -organizing system,

    Kohonen, T., “Automatic formation of topological maps of patterns in a self -organizing system,” Proceedings of 2SCIA, pp. 214-220, 1981a

  12. [20]

    Knowledge Acquisition via Incremental Conceptual Clustering,

    Fisher, D.H., “Knowledge Acquisition via Incremental Conceptual Clustering,” Machine Learning, vol. 2, pp. 139-172, 1987

    1987

  13. [21]

    Bayesian Classification (AutoClass): Theory and Results,

    Cheeseman, P. and Stutz, J., “Bayesian Classification (AutoClass): Theory and Results,” Advances in Knowledge Discovery and Data Mining, Cambridge, 1996

    1996

  14. [22]

    A Fuzzy Relative of the ISODATA Process and Its Use in Detecting Compact Well -Separated Clusters,

    Dunn, J.C., "A Fuzzy Relative of the ISODATA Process and Its Use in Detecting Compact Well -Separated Clusters," Journal of Cybernetics, vol. 3, pp. 32-57, 1973

    1973

  15. [23]

    Pattern Recognition with Fuzzy Objective Function Algorithms,

    Bezdek, J.C., “Pattern Recognition with Fuzzy Objective Function Algorithms,” Plenum Press, New York, 1981

    1981

  16. [24]

    Introduction to Mathematical Optimization: From Linear Programming to Metaheuristics,

    Yang, X.S., “Introduction to Mathematical Optimization: From Linear Programming to Metaheuristics,” Cambridge Int. Science Publishing, 2008

    2008

  17. [25]

    Consensus Clustering and Functional Interpretation of Gene-Expression Data,

    Swift, S., Tucker, A., Vincotti, V., Martin, N., Orengo, C., Liu, X. and Kellam, P., “Consensus Clustering and Functional Interpretation of Gene-Expression Data,” Genome Biology, vol. 5, no. 11, pp. R94.1-R94.16, 2004

    2004

  18. [26]

    Cluster Ensemble and its Applications in Gene Expression Analysis,

    Hu, X. and Yoo, I., "Cluster Ensemble and its Applications in Gene Expression Analysis," Proc. the Second Conference on Asia-Pacific Bioinformatics, vol. 55, pp. 297-302, 1 Jan. 2004

    2004

  19. [27]

    Kaposi's sarcoma-associated herpesvirus-infected primary effusion lymphoma has a plasma cell gene expression profile,

    Jenner, R.G., Maillard, K., Cattini, N., Weiss, R.A., Boshoff, C., Wooster, R. and Kellam, P., “Kaposi's sarcoma-associated herpesvirus-infected primary effusion lymphoma has a plasma cell gene expression profile,” Proc Natl Acad Sci USA, vol. 100, pp. 10399-10404, 2003

    2003

  20. [28]

    Efficiently Mining Gene Expression Data via a Novel Parameterless Clustering Method,

    Tseng, V. S. and Kao, C., “Efficiently Mining Gene Expression Data via a Novel Parameterless Clustering Method,” IEEE/ACM Trans. Comput. Biol. Bioinformatics, vol. 2, no. 4, pp. 355 -365, 2005

    2005

  21. [29]

    Parvin, H., Minaei-Bidgoli, B. (2013). A new imbalanced learning and dictions tree method for breast cancer diagnosis. Journal of Bionanoscience, 7(6), 673-678

    2013

  22. [30]

    Graph-based consensus clustering for class discovery from gene expression data,

    Yu, Z.W., Wong, H.S. and Wang, H.Q., “Graph-based consensus clustering for class discovery from gene expression data,” Bioinformatics, vol. 23, no. 21, pp. 2888-2896, 2007

    2007

  23. [31]

    R., Akbarzadeh, H., Parvin, H., Nejatian, S., Rezaie, V

    Mahmoudi, M. R., Akbarzadeh, H., Parvin, H., Nejatian, S., Rezaie, V. (2021). Consensus function based on cluster-wise two level clustering. Artificial Intelligence Review, 54, 639-665

    2021

  24. [32]

    (2008, November)

    Parvin, H., Alizadeh, H., Moshki, M., Minaei-Bidgoli, B., & Mozayani, N. (2008, November). Divide & conquer classification and optimization by genetic algorithm. In 2008 Third International Conference on Convergence and Hybrid Information Technology (Vol. 2, pp. 858-863). IEEE

    2008

  25. [33]

    Parvin, H., Minaei-Bidgoli, B., & Parvin, S. (2013). A new classifier ensemble methodology based on subspace learning. Journal of Experimental & Theoretical Artificial Intelligence, 25(2), 227-250

    2013

  26. [34]

    Consensus-based Ensembles of Soft Clusterings,

    Punera, K. and Ghosh, J., “Consensus-based Ensembles of Soft Clusterings,” Applied Artificial Intelligence, vol. 22, no. 7-8, pp. 780-810, 2008

    2008

  27. [35]

    An Ensemble Framework for Clustering Protein -Protein Interaction Networks,

    Asur, S., Ucar, D., and Parthasarathy, S., "An Ensemble Framework for Clustering Protein -Protein Interaction Networks," Bioinformatics, vol. 23, no. 13, pp. I29-I40, Jul. 2007

    2007

  28. [36]

    Combining Multiple Clusterings using Evidence Accumulation,

    Fred, A.L.N. and Jain, A.K., "Combining Multiple Clusterings using Evidence Accumulation," IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 27, no. 6, pp. 835 -850, Jun. 2005

    2005

  29. [37]

    A New Efficient Approach in Clustering Ensembles,

    Azimi, J., Abdoos, M. and Analoui, M., “A New Efficient Approach in Clustering Ensembles,” In IDEAL, International Conference on Intelligent Data Engineering and Automated Learning, 2007

    2007

  30. [38]

    Solving Cluster Ensemble Problems by Bipartite Graph Partitioning,

    Fern, X.Z. and Brodley, C.E., "Solving Cluster Ensemble Problems by Bipartite Graph Partitioning," Proc. the Twenty-First International Conference on Machine Learning (ICML '04), vol. 69, ACM Press, New York, NY, pp. 36, 2004

    2004

  31. [39]

    Weighted Cluster Ensemble Using a Kernel Consensus Function,

    Vega-Pons, S., Correa-Morris, J. and Ruiz-Shulcloper, J., “Weighted Cluster Ensemble Using a Kernel Consensus Function,” Progress in Pattern Recognition, Image Analysis and Applications, vol. 5197, pp. 195 - 202, 2008

    2008

  32. [40]

    Advances in Minimum Description Length: Theory and Applications,

    Graunwald, P.D., Myung, I.J. and Pitt, M.A. (Eds.), “Advances in Minimum Description Length: Theory and Applications,” MIT Press, 2004

    2004

  33. [41]

    Parvin, H., Minaei-Bidgoli, B., Punch, W. F. (2013). Data weighing mechanisms for clustering ensembles. Computers & Electrical Engineering, 39(5), 1433-1450

    2013

  34. [42]

    Minaei-Bidgoli, B., Parvin, H., Alizadeh, H., & Punch, W. F. (2014). Effects of resampling method and adaptation on clustering ensemble efficacy. Artificial Intelligence Review, 41, 27 -48

    2014

  35. [43]

    Parvin, H., Seyedaghaee, N., & Parvin, S. (2012). A heuristic scalable classifier ensemble of binary classifier ensembles. Journal of Bioinformatics and Intelligent Control, 1(2), 163-170

    2012

  36. [44]

    Parvin, H., Minaei, B., Alizadeh, H., & Beigi, A. (2011). A novel classifier ensemble method based on class weightening in huge dataset. In Advances in Neural Networks–ISNN 2011: 8th International Symposium on Neural Networks, ISNN 2011, Guilin, China, May 29–June 1, 2011, Pro...

    2011

  37. [45]

    M., Liaw, S

    Hansun, S., Argha, A., Alizadehsani, R., Gorriz, J. M., Liaw, S. T., ... & Marks, G. B. (2024). A New Ensemble Transfer Learning Approach With Rejection Mechanism for Tuberculosis Disease Detection. IEEE Transactions on Radiation and Plasma Medical Sciences

    2024

  38. [46]

    Sadeghi, A., Hajati, F., Rezaee, A., Sadeghi, M., Argha, A. (2024). 3DECG-Net: ECG fusion network for multi-label cardiac arrhythmia detection. Computers in Biology and Medicine, 182, 109126

    2024

  39. [47]

    Niu, H., et al. (2020). Deep feature learnt by conventional deep neural network. Computers & Electrical Engineering, 84, 106656

    2020

  40. [48]

    H., Hossein Nasirpour, M., Mirshekari Chargari, M., Mohammadi, F., Moravvej, S

    Hong, L., Modirrousta, M. H., Hossein Nasirpour, M., Mirshekari Chargari, M., Mohammadi, F., Moravvej, S. V., Nahavandi, S. (2023). GAN‐LSTM‐3D: An efficient method for lung tumour 3D reconstruction enhanced by attention‐based LSTM. CAAI Transactions on Intelligence Technology

    2023

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.