Pith. sign in

REVIEW 4 major objections 5 minor 50 references

Fair Bayesian Model-Based Clustering

T0 review · 4 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read Fair Bayesian clustering infers K and keeps sensitive groups balanced by encoding fairness through matching maps instead of hard constraints.

desk verdict FBC is a genuine step forward for Bayesian fair clustering, but its real-data claims rest on an unproven r>0 heuristic and single-draw experiments. read the letter →

arxiv 2506.12839 v1 pith:TKXUKIHG submitted 2025-06-15 stat.ML cs.AIcs.LG

Jihu Lee , Kunwoong Kim , Yongdai Kim This is my paper · ORCID
classification stat.MLcs.AIcs.LG MSC 62F1562H30
keywords fairclusteringBayesianmixturemodelmodel-basedgroupfairnessmatchingmapunknownnumberofclustersMCMCcategoricaldata
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

Most fair clustering methods are built on K-means, so they need the number of clusters and a distance function supplied in advance. This paper proposes Fair Bayesian Clustering (FBC), a mixture-model approach that treats the number of clusters as an unknown parameter and enforces group fairness through the prior rather than through constraints. The key move is to construct fair cluster assignments by matching instances across sensitive groups, so that each matched pair receives the same cluster label. Because the prior is parameterized by freely choosable matching maps, posterior inference by MCMC is straightforward, and the method applies to any likelihood, including categorical data. The paper reports that FBC infers cluster counts reasonably and matches the utility-fairness trade-off of existing fair clustering baselines.

What carries the argument

The load-bearing object is the matching map $T:[n_1]\to[n_0]$ with $|T^{-1}(i)|=\beta$, together with the masked subset $E$ of size $m$ and an arbitrary map $T_0$ for $E$. Conditional on $T$, each group-1 instance $j$ is assigned the cluster of its match, $Z_j^{(1)}=Z_{T(j)}^{(0)}$, and instances in $E$ receive independent labels via $T_0$. This construction guarantees that the support of the assignment distribution lies in $\mathcal{Z}^{\mathrm{Fair}}_{m/n_1}$ when $r=0$, and the set of all matching maps covers every perfectly fair assignment. An energy-based prior $e(T)=\exp\big(-\sum_j D(X_{T(j)}^{(0)},X_j^{(1)})/(n_1\tau)\big)$ favors matches between similar instances, and the MCMC algorithm alternately samples the matching structure $(T,T_0,E)$ and the mixture partition $C$ of $[n_0]$ using a modified Dirichlet-process-mixture sampler.

What would settle it

Run FBC with $m=0$ on synthetic data with $n_0=60$, $n_1=90$ ($\beta=1$, $r=30$) where the smaller group contains three well-separated clusters of 20 points each but the chosen subset $R$ contains 20 points from the first cluster and 5 from each of the other two. Because the sufficient condition $|C_k^{(0)}|/n_0=|C_k^{(0)}\cap R|/r$ fails, the posterior's matching assignments can make the group-1 proportion in the first cluster far exceed the group-0 proportion, so the reported $\Delta(Z)$ would be bounded well above zero; this would show the $r>0$ heuristic does not deliver the fairness guarantee that the $r=0$ theory provides.

Watch

Extended reading notes

Core claim

The paper's central claim is that group-fair clustering with an unknown number of clusters can be formulated as a Bayesian mixture model whose prior is supported on fair assignments. A cluster assignment $Z$ is perfectly fair ($\Delta(Z)=0$) exactly when one can find a matching map $T$ from the larger sensitive group to the smaller one such that every group-1 instance inherits the cluster label of its matched group-0 instance, provided $n_1=\beta n_0$ (Theorem 3.2). This equivalence turns fairness from a hard constraint into a reparameterization: the model draws cluster labels for group 0 from a categorical distribution and copies them along matching maps, so no constrained parameter space is needed. A masked set $E$ of $m$ group-1 instances, whose labels are drawn independently, relaxes perfect fairness to $\Delta(Z)\le m/n_1$ (Theorem A.3), giving a tunable fairness knob. On DIABETES, ADULT, and BANK, FBC reports the lowest $\Delta$ among compared methods at comparable cost, infers $K$ through its posterior, and improves fairness on binarized categorical data.

Load-bearing premise

When the two sensitive groups have sizes $n_1=\beta n_0+r$ with $r>0$, the paper replaces exact fairness with a heuristic: fix a subset $R$ of size $r$ and declare assignments fair if they arise from a matching map that pairs group-1 points to group-0 points in a balanced way, with the $r$ 'extra' pairs concentrated on $R$. The load-bearing premise is that a random (or cluster-centered) $R$ represents the smaller group's distribution well, because all real datasets in the experiments have $r>0$ while the proven fairness bound applies only to $r=0$.

Editorial extensions

If this is right

  • FBC infers the number of clusters from data, removing the need to fix $K$ in advance.
  • Because fairness is encoded in the prior, FBC applies to any data type with a defined likelihood, including categorical data modeled by Bernoulli mixtures.
  • With $r=0$, setting the masked-set size $m$ controls the fairness level: the resulting assignment satisfies $\Delta(Z)\le m/n_1$.
  • On real benchmark datasets, FBC achieves a utility-fairness trade-off comparable to (and often fairer than) SFC, VFC, and Fair MFM.
  • The matching-map formulation extends to three or more sensitive groups, with a bound $\Delta(Z)\le (m_1/n_1+m_2/n_2)/2$.

Reading between the lines

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

  • The reparameterization suggests that other constrained clustering tasks, such as diversity or capacity constraints, could be encoded by modifying the matching class instead of adding penalty terms.
  • The paper leaves assignment of new data as future work; a natural extension is to learn a parametric approximation of the matching map and apply it to test points.
  • The $r>0$ heuristic deserves a theoretical worst-case analysis: if $R$ is adversarially chosen, the fairness violation of assignments in $\mathcal{T}_R$ can be large, so a data-dependent criterion for choosing $R$ would strengthen the method.
  • Since the prior over $T$ uses an energy based on a pre-specified distance, FBC inherits the need for a meaningful distance for the matching step even though the likelihood itself can be model-based; a fully distance-free variant could be explored.
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

4 major / 5 minor

Summary. The paper proposes Fair Bayesian Clustering (FBC), a Bayesian mixture model with an unknown number of clusters that enforces group fairness through a matching-map prior. For balanced groups (n0 = n1 or n1 = beta n0), the support of the proposed prior is exactly the set of perfectly fair assignments (Theorems 3.1 and 3.2). For unequal groups with remainder r > 0, fairness is defined through a heuristic subset R, and an epsilon-fair version masks m instances and is claimed to achieve Delta <= m/n1. An MCMC algorithm is developed and experiments on toy and real datasets (DIABETES, ADULT, BANK) report competitive utility-fairness trade-offs, inferred numbers of clusters, and categorical-data performance.

Significance. If the r = 0 theory and the r > 0 heuristic were rigorously supported, FBC would be a useful contribution: it extends fair clustering beyond K-means to likelihood-based models with unknown K and categorical data, and the matching construction is elegant and computationally attractive. The exact-support results for r = 0 are clean and correctly proved. However, the headline claims go beyond the proven theory because all real-data experiments use r > 0 and are governed by an unproven approximation rather than by Theorem A.3. The manuscript also reports all experimental numbers from a single randomly selected posterior sample, without repeated runs or error bars.

major comments (4)
  1. [Section 3.3.2 and Theorem A.3] The fairness guarantee for the actual method is not established for any of the real datasets. For n1 = beta n0 + r with r > 0, the prior's support is {Z : Z_j^(1) = Z_T(j)^(0) for some T in T_R}; such Z need not lie in ZFair_0, and for a fixed T in T_R the fairness level equals Delta(Z) = r/(2 n1) * sum_k |p_k - q_k|, where p_k = |C_k^(0)|/n0 and q_k = |C_k^(0) intersect R|/r. This can be as large as r/n1 on the support of the prior (e.g., about 0.35 for BANK). The only rigorous bound, Theorem A.3, assumes r = 0, and with m = 0 it would give Delta = 0 only in that case. For DIABETES, ADULT, and BANK the remainders are r = 24, 82, and 212, so the reported Delta values (0.012, 0.006, 0.020) are not covered by any theorem; the abstract's statement that the prior 'puts its mass only on fair clusters' is therefore unsupported for the experimental setting. Please either prove a rigorous deterministic or high-probability bound on Delta for random R in the r > 0 case, or restrict the fairness claim to r = 0 and describe the r > 0 construction as a heuristic.
  2. [Section A, proof of Theorem A.1] The proof that a fair Z admits a map T in T is not valid as written. After defining k* = argmin_k |C_k^(0)|, the proof asserts a_l = |C_l^(0)| / |C_k^(0)| in N for all l; this is false for fair assignments, e.g., n0 = 15, n1 = 20 (beta = 1, r = 5) with |C_1^(0)| = 6 and |C_2^(0)| = 9 gives a_2 = 1.5, while the assignment is fair since |C_1^(1)| = 8 and |C_2^(1)| = 12. Thus the existence claim for r > 0 is unproven. Please repair the construction (e.g., define alpha_l = r |C_l^(0)| / n0 directly, which is an integer for fair Z) or state a corrected theorem.
  3. [Section 6.1 and Tables 1-3] All headline numbers are from one randomly chosen post-burn-in sample, with no repeated runs or measures of posterior variability. The utility-fairness trade-off comparisons and the ablation conclusions in Tables 1, 2, 3, 4, and 5 could therefore be driven by a single draw. Please report means and standard deviations over repeated MCMC runs or over multiple posterior samples, and state the selection protocol explicitly.
  4. [Section C, Theorem C.1] The multinary extension is stated without the balance assumptions needed by Theorem A.3. The proof of Theorem C.1 invokes Theorem A.3 pairwise, but that theorem requires r = 0 for each pair; the BANK experiment with three sensitive groups has sizes (586, 305, 109), so the bound Delta <= 1/2 (m1/n1 + m2/n2) is not proven for that experiment. Please add explicit hypotheses (e.g., exact divisibility for all pairs) or supply a proof for unequal group sizes.
minor comments (5)
  1. [Section 3.3.2] The set T of matching maps for the r > 0 case is defined similarly to the r = 0 case but with a different cardinality condition; please use distinct notation (e.g., T_r) to avoid confusion.
  2. [Throughout] Theorems 3.1 and 3.2 are called Propositions in the main text, and Theorems A.1, A.2, and A.3 are inconsistently labeled as Propositions in the appendix. Please unify the theorem/proposition numbering.
  3. [Section 4.2] The prior for E is described as uniform on [n1 : m], which is nonstandard notation; it should say the collection of all subsets of [n1] of cardinality m.
  4. [Section 4.2] Definition 4.1 is referred to as 'Theorem 4.1' in the text; please correct this cross-reference.
  5. [Section D.3.2] In the test-data evaluation, the optimal transport maps T_test and T* are used without specifying the cost function or the algorithm; please give the implementation details.

Circularity Check

1 steps flagged · score 2.0 of 10

FBC's fairness guarantee is a self-contained prior construction; only mild definitional circularity in the r>0 case, where 'fair' is redefined via T_R.

  1. self definitional [Section 3.3.2, Case of r>0; cf. Abstract and Theorem A.3]
    "To resolve this difficulty, we propose a heuristic modification of the definition of 'fairness of Z'. Let R be a subset of [n0] with |R|=r, and let TR be a subset of T such that RT =R. Then, we say that Z is fair if there exists T∈ TR such that Z(1)j =Z(0)T(j). Note that a fair Z may not belong to ZFair0 but the violation of fairness would be small when |C(0)k|/n0 ≈ |C(0)k∩R|/r."

    For r>0, the prior's support is fair only under the redefined notion (existence of T∈TR), not under the original ∆(Z) measure of Eq. (7) used in Tables 1-3 and in the abstract's claim that the prior 'puts its mass only on fair clusters'. The proved fairness bound (Theorem A.3) requires r=0, so for real datasets the claimed fairness guarantee reduces to a definitional convention plus an unproven approximation about the subset R. The fairness-control experiment then confirms this construction rather than an independent prediction.

full rationale

The central construction is not circular: the matching-map prior is explicitly designed so that its support equals ZFair0 when n1=βn0, and Propositions 3.1-3.2 prove that equivalence from the definition of ∆; Theorem A.3 derives the ε=m/n1 bound from the masked-set mechanism for r=0. These are self-contained mathematical proofs, and FBC is benchmarked against external SFC, VFC, and MFM baselines. The self-citations [10,11] are contextual acknowledgments of the matching idea, not load-bearing evidence for the theorems. The only mild definitional element is the r>0 case, where Section 3.3.2 explicitly redefines 'fair' via T_R and relies on a heuristic approximation instead of the original ∆; this is acknowledged and empirically checked, so it is a caveat or correctness risk rather than a forced equivalence. Hence a low circularity score of 2.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The method's central construction is free of fitted parameters: the fairness guarantee is built into the prior. The main burden falls on the heuristic r>0 extension, which assumes a representative subset R, and on standard Bayesian modeling assumptions. The hyperparameters (tau, m, gamma, kappa, a, b) are chosen by hand; the paper argues via ablations that results are insensitive to tau and kappa.

free parameters (5)
  • tau (temperature in energy prior) = 1.0 (default; ablation over 0.1, 1.0, 10.0)
    Controls the strength of the prior on matching similar points. Chosen by hand, not fitted; ablation in Section D.3.4 shows low sensitivity.
  • m (size of masked set E) = 0 for perfect fairness in main results
    User-specified fairness level. Theoretic bound Delta <= m/n1 holds for r=0. This is the main knob for the utility-fairness trade-off.
  • gamma (Dirichlet concentration) = 1
    Prior parameter for cluster proportions, following Miller and Harrison's MFM.
  • kappa (Geometric prior on K) = 0.1
    Prior parameter for number of clusters; sensitivity analysis in Figure 8.
  • a, b (NIW hyperparameters) = a=b=1
    Prior parameters for Gaussian components in the conjugate case.
assumptions (4)
  • domain assumption The likelihood f is correctly specified for the data (Gaussian for continuous, product Bernoulli for categorical).
    Model definition in Section 4.1 assumes a known parametric family; misspecification could bias the clusters.
  • domain assumption The sensitive attribute is known and binary, with a natural reference group 0.
    The construction matches group 1 to group 0; an extension for multinary attributes is given in Section C.
  • ad hoc to paper For r>0, a random subset R of [n0] of size r has an empirical distribution close to that of [n0], so the modified fairness definition is approximately exact.
    Section 3.3.2 relies on this heuristic; no theorem establishes it, and it is the load-bearing assumption for all real datasets.
  • domain assumption The MCMC sampler converges to the posterior within the 1200 iterations used.
    Supported empirically by autocorrelation and trace plots in Section D.3.4, but without theoretical guarantee.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Fair Bayesian Model-Based Clustering." pith.science (2026). https://pith.science/paper/TKXUKIHG

@misc{pith2026250612839,
  author       = {Pith},
  title        = {Pith review of: Fair Bayesian Model-Based Clustering},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/TKXUKIHG}},
  note         = {Machine review of arXiv:2506.12839}
}
abstract

Fair clustering has become a socially significant task with the advancement of machine learning technologies and the growing demand for trustworthy AI. Group fairness ensures that the proportions of each sensitive group are similar in all clusters. Most existing group-fair clustering methods are based on the $K$-means clustering and thus require the distance between instances and the number of clusters to be given in advance. To resolve this limitation, we propose a fair Bayesian model-based clustering called Fair Bayesian Clustering (FBC). We develop a specially designed prior which puts its mass only on fair clusters, and implement an efficient MCMC algorithm. Advantages of FBC are that it can infer the number of clusters and can be applied to any data type as long as the likelihood is defined (e.g., categorical data). Experiments on real-world datasets show that FBC (i) reasonably infers the number of clusters, (ii) achieves a competitive utility-fairness trade-off compared to existing fair clustering methods, and (iii) performs well on categorical data.

Figures

Figures reproduced from arXiv: 2506.12839 by the authors.

Figure 1
Figure 1. An example illustration of our fair assignment when [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Visualization of the clustering results on a toy dataset (left - MFM with [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
Figure 3
Figure 3. draws the posterior distributions of K for the three datasets. The posterior distributions are well-concentrated around the posterior modes. See [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: (Top three) Trade-off between m (the size of E) and the fairness level ∆. Smaller ∆, fairer the clustering. (Bottom three) Trade-off between m (the size of E) and the Bal. Larger Bal, fairer the clustering. Analysis with a multinary sensitive attribute (BANK) We analyz…
Figure 5
Figure 5. Figure 5: Trade-off between the size of E and the fairness level ∆. Smaller ∆, fairer the clustering. The x-axis is 1 − m/nmax where nmax = max(n0, n1, n2) = max(109, 305, 586) = 586. 21 [PITH_FULL_IMAGE:figures/full_fig_p021_5.png]
Figure 6
Figure 6. Figure 6: A visualization of the construction of fair assignments [PITH_FULL_IMAGE:figures/full_fig_p022_6.png]
Figure 7
Figure 7. Figure 7: Number of clusters K vs. Negative log-likelihood (NLL) on test data. 22 [PITH_FULL_IMAGE:figures/full_fig_p022_7.png]
Figure 8
Figure 8. Figure 8: below shows the results, suggesting that FBC is not sensitive to the choice of κ. For example, the posterior mode of K is consistently 3 for ADULT dataset [PITH_FULL_IMAGE:figures/full_fig_p025_8.png]
Figure 9
Figure 9. Figure 9: h vs. Autocorrelation functions for (left) DIABETES, (center) ADULT, and (right) BANK datasets. NLL [PITH_FULL_IMAGE:figures/full_fig_p026_9.png]
Figure 10
Figure 10. Figure 10: Trace plots of NLL on (left) DIABETES, (center) ADULT, and (right) BANK datasets. 26 [PITH_FULL_IMAGE:figures/full_fig_p026_10.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

50 extracted references · 47 canonical work pages

  1. [1]

    Fair clustering through fairlets.Advances in neural information processing systems, 30, 2017

    Flavio Chierichetti, Ravi Kumar, Silvio Lattanzi, and Sergei Vassilvitskii. Fair clustering through fairlets.Advances in neural information processing systems, 30, 2017

    work page 2017

  2. [2]

    Sylvia Richardson and Peter J Green. On bayesian analysis of mixtures with an unknown number of components (with discussion).Journal of the Royal Statistical Society Series B: Statistical Methodology, 59(4):731–792, 1997

    work page 1997

  3. [3]

    Bayesian finite mixtures with an unknown number of components: The allocation sampler.Statistics and Computing, 17:147–162, 2007

    Agostino Nobile and Alastair T Fearnside. Bayesian finite mixtures with an unknown number of components: The allocation sampler.Statistics and Computing, 17:147–162, 2007

    work page 2007

  4. [4]

    How many clusters?Bayesian Analysis, 3(1):101–120, 2008

    Peter McCullagh and Jie Yang. How many clusters?Bayesian Analysis, 3(1):101–120, 2008

    work page 2008

  5. [5]

    Mixture models with a prior on the number of components.Journal of the American Statistical Association, 113(521):340–356, 2018

    Jeffrey W Miller and Matthew T Harrison. Mixture models with a prior on the number of components.Journal of the American Statistical Association, 113(521):340–356, 2018

    work page 2018

  6. [6]

    A bayesian analysis of some nonparametric problems.The annals of statistics, pages 209–230, 1973

    Thomas S Ferguson. A bayesian analysis of some nonparametric problems.The annals of statistics, pages 209–230, 1973

    work page 1973

  7. [7]

    Mixtures of dirichlet processes with applications to bayesian nonparametric problems.The annals of statistics, pages 1152–1174, 1974

    Charles E Antoniak. Mixtures of dirichlet processes with applications to bayesian nonparametric problems.The annals of statistics, pages 1152–1174, 1974

    work page 1974

  8. [8]

    Bayesian density estimation and inference using mixtures

    Michael D Escobar and Mike West. Bayesian density estimation and inference using mixtures. Journal of the american statistical association, 90(430):577–588, 1995

    work page 1995

Show all 50 references
  1. [9]

    Markov chain sampling methods for dirichlet process mixture models.Journal of computational and graphical statistics, 9(2):249–265, 2000

    Radford M Neal. Markov chain sampling methods for dirichlet process mixture models.Journal of computational and graphical statistics, 9(2):249–265, 2000

    2000

  2. [10]

    Fairness through matching.Transactions on Machine Learning Research, 2025

    Kunwoong Kim, Insung Kong, Jongjin Lee, Minwoo Chae, Sangchul Park, and Yongdai Kim. Fairness through matching.Transactions on Machine Learning Research, 2025

    2025

  3. [11]

    Fair clustering via alignment, 2025

    Kunwoong Kim, Jihu Lee, Sangchul Park, and Yongdai Kim. Fair clustering via alignment, 2025

    2025

  4. [12]

    Gaussian mixture clustering and imputation of microarray data.Bioinformatics, 20(6):917–923, 2004

    Ming Ouyang, William J Welsh, and Panos Georgopoulos. Gaussian mixture clustering and imputation of microarray data.Bioinformatics, 20(6):917–923, 2004

    2004

  5. [13]

    Gaussian mixture models.Encyclopedia of biometrics, 741(659-663), 2009

    Douglas A Reynolds et al. Gaussian mixture models.Encyclopedia of biometrics, 741(659-663), 2009

    2009

  6. [14]

    Variable selection for clustering with gaussian mixture models.Biometrics, 65(3):701–709, 2009

    Cathy Maugis, Gilles Celeux, and Marie-Laure Martin-Magniette. Variable selection for clustering with gaussian mixture models.Biometrics, 65(3):701–709, 2009

    2009

  7. [15]

    A robust em clustering algorithm for gaussian mixture models.Pattern Recognition, 45(11):3950–3961, 2012

    Miin-Shen Yang, Chien-Yo Lai, and Chih-Ying Lin. A robust em clustering algorithm for gaussian mixture models.Pattern Recognition, 45(11):3950–3961, 2012

    2012

  8. [16]

    Gaussian mixture model clustering with incomplete data

    Yi Zhang, Miaomiao Li, Siwei Wang, Sisi Dai, Lei Luo, En Zhu, Huiying Xu, Xinzhong Zhu, Chaoyun Yao, and Haoran Zhou. Gaussian mixture model clustering with incomplete data. ACM Transactions on Multimedia Computing, Communications, and Applications (TOMM), 17(1s):1–14, 2021

    2021

  9. [17]

    Robust mixture modelling using the t distribution

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

    2000

  10. [18]

    Finite mixture modelling using the skew normal distribution.Statistica Sinica, pages 909–927, 2007

    Tsung I Lin, Jack C Lee, and Shu Y Yen. Finite mixture modelling using the skew normal distribution.Statistica Sinica, pages 909–927, 2007

    2007

  11. [19]

    Bayesian inferences of latent class models with an unknown number of classes.Psychometrika, 79(4):621–646, 2014

    Jia-Chiun Pan and Guan-Hua Huang. Bayesian inferences of latent class models with an unknown number of classes.Psychometrika, 79(4):621–646, 2014

    2014

  12. [20]

    John Wiley & Sons, 2000

    Geoffrey J McLachlan and David Peel.Finite mixture models. John Wiley & Sons, 2000

    2000

  13. [21]

    A split-merge markov chain monte carlo procedure for the dirichlet process mixture model.Journal of computational and Graphical Statistics, 13(1):158– 182, 2004

    Sonia Jain and Radford M Neal. A split-merge markov chain monte carlo procedure for the dirichlet process mixture model.Journal of computational and Graphical Statistics, 13(1):158– 182, 2004. 10

    2004

  14. [22]

    Scalable fair clustering

    Arturs Backurs, Piotr Indyk, Krzysztof Onak, Baruch Schieber, Ali Vakilian, and Tal Wagner. Scalable fair clustering. InInternational Conference on Machine Learning, pages 405–413. PMLR, 2019

    2019

  15. [23]

    Guarantees for spectral clustering with fairness constraints

    Matthäus Kleindessner, Samira Samadi, Pranjal Awasthi, and Jamie Morgenstern. Guarantees for spectral clustering with fairness constraints. InInternational conference on machine learning, pages 3458–3467. PMLR, 2019

    2019

  16. [24]

    Variational fair clustering

    Imtiaz Masud Ziko, Jing Yuan, Eric Granger, and Ismail Ben Ayed. Variational fair clustering. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 35, pages 11202–11209, 2021

    2021

  17. [25]

    Deep fair clustering for visual learning

    Peizhao Li, Han Zhao, and Hongfu Liu. Deep fair clustering for visual learning. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 9070–9079, 2020

    2020

  18. [26]

    Deep fair clustering via maximizing and minimizing mutual information: Theory, algorithm and metric

    Pengxin Zeng, Yunfan Li, Peng Hu, Dezhong Peng, Jiancheng Lv, and Xi Peng. Deep fair clustering via maximizing and minimizing mutual information: Theory, algorithm and metric. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 23986–23995, 2023

    2023

  19. [27]

    Fair algorithms for clustering.Advances in Neural Information Processing Systems, 32, 2019

    Suman Bera, Deeparnab Chakrabarty, Nicolas Flores, and Maryam Negahbani. Fair algorithms for clustering.Advances in Neural Information Processing Systems, 32, 2019

    2019

  20. [28]

    Kfc: A scalable approximation algorithm for kcenter fair clustering

    Elfarouk Harb and Ho Shan Lam. Kfc: A scalable approximation algorithm for kcenter fair clustering. In H. Larochelle, M. Ranzato, R. Hadsell, M.F. Balcan, and H. Lin, editors,Advances in Neural Information Processing Systems, volume 33, pages 14509–14519. Curran Associates, Inc., 2020

    2020

  21. [29]

    Efficient sampling for bipartite matching problems

    Maksims V olkovs and Richard Zemel. Efficient sampling for bipartite matching problems. In F. Pereira, C.J. Burges, L. Bottou, and K.Q. Weinberger, editors,Advances in Neural Information Processing Systems, volume 25. Curran Associates, Inc., 2012

    2012

  22. [30]

    Estimating mixture of dirichlet process models.Journal of Computational and Graphical Statistics, 7(2):223–238, 1998

    Steven N MacEachern and Peter Müller. Estimating mixture of dirichlet process models.Journal of Computational and Graphical Statistics, 7(2):223–238, 1998

    1998

  23. [31]

    Everhart, W

    Jack Smith, J. Everhart, W. Dickson, W. Knowler, and Richard Johannes. Using the adap learning algorithm to forcast the onset of diabetes mellitus.Proceedings - Annual Symposium on Computer Applications in Medical Care, 10, 11 1988

    1988

  24. [32]

    Barry Becker and Ronny Kohavi. Adult. UCI Machine Learning Repository, 1996. DOI: https://doi.org/10.24432/C5XW20

    1996 doi

  25. [33]

    A data-driven approach to predict the success of bank telemarketing.Decision Support Systems, 62:22–31, 2014

    Sérgio Moro, Paulo Cortez, and Paulo Rita. A data-driven approach to predict the success of bank telemarketing.Decision Support Systems, 62:22–31, 2014

    2014

  26. [34]

    Fair clustering under a bounded cost.Advances in Neural Information Processing Systems, 34:14345–14357, 2021

    Seyed Esmaeili, Brian Brubach, Aravind Srinivasan, and John Dickerson. Fair clustering under a bounded cost.Advances in Neural Information Processing Systems, 34:14345–14357, 2021

    2021

  27. [35]

    Error detecting and error correcting codes.The Bell system technical journal, 29(2):147–160, 1950

    Richard W Hamming. Error detecting and error correcting codes.The Bell system technical journal, 29(2):147–160, 1950

    1950

  28. [36]

    Extensions to the k-means algorithm for clustering large data sets with categori- cal values.Data mining and knowledge discovery, 2(3):283–304, 1998

    Zhexue Huang. Extensions to the k-means algorithm for clustering large data sets with categori- cal values.Data mining and knowledge discovery, 2(3):283–304, 1998

    1998

  29. [37]

    Clustering categorical data based on distance vectors.Journal of the American Statistical Association, 101(473):355–367, 2006

    Peng Zhang, Xiaogang Wang, and Peter X-K Song. Clustering categorical data based on distance vectors.Journal of the American Statistical Association, 101(473):355–367, 2006. 11 A Theoretical results for Section 3 Theorem 3.1Assume that n0 =n 1 = ¯n.Then, we have: Z∈ ZFair 0 ⇐ ...

    2006

  30. [38]

    InitializeC={[n 0]}(i.e., a single cluster)

  31. [39]

    For i= 1,

    Repeat the following steps N times, to obtain N samples. For i= 1, . . . , n0: Remove element i∈[n 0] and its matched elements in J({i};T,T 0, E) :={j∈E:T 0(j) = i} ∪ {j∈[n1]\E:T(j) =i}fromC. Then, place them • toc ′ ∈ C \iwith probability ∝(|c ′|+γ) m(X c′ ∪X {i}|T,T 0, E) m(...

  32. [40]

    InitializeC={[n 0]}(i.e., a single cluster) withϕ [n0] ∼H

  33. [41]

    For i= 1,

    Repeat the following steps N times, to obtain N samples. For i= 1, . . . , n0: Remove element i∈[n 0] and its matched elements in J({i}) from C. Then, generate m independent auxiliary variablesϕ (1), . . . , ϕ(m) ∼H. Compute the assignment weights as: wc′ = (|c′|+γ) Y x∈X {i} ...

  34. [42]

    Toy dataset: We build a 2D toy dataset from a 6-component Gaussian mixture model with unit covariance matrix I2. For D(0), we draw 600 samples from each N([−5,−30],I 2)/3 + N([−5,0],I 2)/3 +N([−5,30],I 2)/3 and Similarly for D(1), we draw 600 samples from N([−5,−29.5],I 2)/3 +...

  35. [43]

    The sample size is 768 (sensitive group sizes: 396 and 372)

    DIABETES: The diabetes dataset is a collection of data spanning five years, consisting of various physical indicators (e.g., glucose concentration, blood pressure, BMI, etc., totaling 7 features) of Pima Indian women*. The sample size is 768 (sensitive group sizes: 396 and 372...

  36. [44]

    It is extracted from 1994 U.S

    ADULT: The adult income dataset is a collection of data consisting of several demographic features including employment features. It is extracted from 1994 U.S. Census database [32]. We subsample 1,000 data points (sensitive group sizes: 694 and 306) from the original dataset....

    1994

  37. [45]

    We use 6 continuous features (age, call duration, 3-month Euribor rate, number of employees, con- sumer price index, and number of contacts during the campaign)

    BANK: The bank marketing dataset is a collection of data from a Portuguese bank’s direct marketing campaigns, each corresponding to an individual client contacted [33]. We use 6 continuous features (age, call duration, 3-month Euribor rate, number of employees, con- sumer pric...

  38. [46]

    That is, |Dtest 0 |= |Dtest 1 |=n test

    We prepare test dataset Dtest 0 and Dtest 1 , each containing ntest samples. That is, |Dtest 0 |= |Dtest 1 |=n test

  39. [47]

    , Z(0),test ntest , Z(1),test 1 ,

    To construct fair assignments Ztest := (Z (0),test 1 , . . . , Z(0),test ntest , Z(1),test 1 , . . . , Z(1),test ntest ) on Dtest 0 ∪ Dtest 1 , we build a one-to-one matching map between Dtest 0 and Dtest 1 , using the optimal transport mapT test

  40. [48]

    We match instances inD test 0 to those inD 0 by using an optimal transportT ∗

  41. [49]

    We assignZ (0),test j =Z (0) T∗(j) andZ (1),test j =Z (0),test Ttest(j) forj∈[n test]

  42. [50]

    See Figure 6 for a simple visualization of this construction whenn test = 4

    For each posterior sample obtained by FBC, we calculate the log-likelihood of test data conditional onZ test. See Figure 6 for a simple visualization of this construction whenn test = 4. Figure 6: A visualization of the construction of fair assignments Ztest for the test data ...

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.