Pith. sign in

REVIEW 4 major objections 3 minor 57 references

Label-shift robust federated feature screening for high-dimensional classification

T0 review · 4 major / 3 minor · reviewed 2026-08-07 · deepseek-v4-flash

Pith's one-line read The paper introduces LR-FFS, a federated feature screening utility whose class-proportion-free estimator matches centralized sure screening rates under label shift.

desk verdict A genuinely new screening utility and a clean federated estimator, but the sure-screening theorem is proved for a target that under arbitrary label shift differs from the pooled utility. read the letter →

arxiv 2506.00379 v1 pith:2N4IICM5 submitted 2025-05-31 stat.ML cs.LGstat.ME

classification stat.MLcs.LGstat.ME
keywords federatedlearningfeaturescreeninglabelshifthigh-dimensionalclassificationMann-Whitneystatisticsurefalsediscoveryratemodel-free
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 settle a specific failure mode in distributed high-dimensional classification: when the same features predict the same classes but class prevalence differs across clients (label shift), existing screening utilities estimate different quantities on different clients and drift from the pooled ranking. The authors propose LR-FFS, a utility defined as the largest, over classes $y_r$, of the deviation $|E_{Y=y_r}(F_{Y\neq y_r}(X_j))-1/2|$, a Mann-Whitney comparison of class $y_r$ against all other classes. They give a one-shot federated estimator that corrects class-proportion bias through a global $\pi_r^*$ equation, then aggregates per-client Mann-Whitney estimates. The main theoretical result is the sure screening property: with probability at least $1-6sR\exp(-c_5N^{1-2\kappa-4\xi})$ the estimated active set contains all $s$ relevant features, at the same rate as centralized screening. If correct, federated systems can screen features before modeling, with no raw data leaving clients and no extra communication or local computation relative to existing baselines.

What carries the argument

The load-bearing object is the class-conditional Mann-Whitney identity $\gamma_{j,r}=E_{Y=y_r}(F_{Y\neq y_r}(X_j))=P(X_{j,i_1}<X_{j,i_2}\mid Y_{i_1}\neq y_r, Y_{i_2}=y_r)$, the probability that a class-$y_r$ observation exceeds a non-$y_r$ observation on feature $j$. The screening utility is $\omega_{j,r}=|\gamma_{j,r}-1/2|$, and LR-FFS takes the maximum over classes, so the class weight is an indicator rather than a proportion-dependent coefficient. The distributed estimator decomposes $\gamma_{j,r}=U_{j,r}/\theta_r$ with $\theta_r=\pi_r(1-\pi_r)$, estimates $U_{j,r}$ and $\theta_r$ by local U-statistics weighted with effective sample sizes $h_l=\lfloor n_l/2\rfloor$, and removes label-shift bias by replacing $\pi_r$ with the global solution $\pi_r^*$ of $\sum_l h_l\pi_r^l(1-\pi_r^l)=\sum_l h_l\pi_r^*(1-\pi_r^*)$. This correction is what lets every client estimate the same $\gamma_{j,r}$ from summary statistics alone.

What would settle it

Run a three-client, three-class simulation with identical $P(X\mid Y)$ and fixed share of class 1, but make client A see only class 2 as 'not class 1' and client B see only class 3 as 'not class 1', then compare the federated LR-FFS estimate of $\omega_j$ for a feature that separates classes 2 and 3 with the same statistic computed on the pooled data; a material gap shows the invariance claim does not extend to arbitrary label shift.

Watch

Extended reading notes

Core claim

The central discovery is a label-shift-invariant screening target. Write $\omega_{j,r}=|E_{Y=y_r}(F_{Y\neq y_r}(X_j))-1/2|$ for the conditional Mann-Whitney contrast of feature $X_j$ between class $y_r$ and the remaining classes; LR-FFS takes $\omega_j=\max_r\omega_{j,r}$. Because the maximum is taken over class-specific contrasts and the weights do not depend on class proportions, the target utility is identical for every client, whereas CRU, MV-SIS, and CAVS are shown to be special cases of a common framework whose weights are class-proportion functions and therefore shift under label skew. Under the setting that $P(X\mid Y)$ is shared across clients while $P(Y)$ varies, the federated estimator $\bar\omega_j$ obtained by bias-corrected weighted U-statistics has $\mathrm{MSE}=O(N^{4\xi-1})$, satisfies the sure screening bound of Theorem 11, the ranking consistency of Theorem 12, and the FDR control of Theorems 13--14. The claim is that screening effectiveness is maintained at the level of centralized processing without extra computational cost, and the numerical experiments are presented as supporting evidence across normal, heavy-tailed, and missing-category settings.

Load-bearing premise

The argument assumes that all clients end up estimating one common target utility; this is guaranteed for the focal class's share, but if the relative proportions of the other classes differ across clients, the mixture of those other classes differs by client and the common-target premise can fail.

Editorial extensions

If this is right

  • Under the sure screening property, with the threshold $\delta=cN^{-\eta}$, all relevant features are retained with probability at least $1-6sR\exp(-c_5N^{1-2\kappa-4\xi})$, matching the rate of a single machine that sees all $N$ samples.
  • The federated protocol transmits only $mR(p+1)$ summary statistics and costs $O(n_l^2 p)$ per client locally, so label-shift correction adds no communication or computational burden relative to existing distributed screening.
  • When active and inactive utilities are separated by a gap of order $N^{-\eta}$, relevant features are uniformly ranked above irrelevant ones with probability tending to one, so a threshold exists that perfectly separates them.
  • The permutation-based FDR procedure controls the false discovery rate at level $\alpha+o(1)$ in high-dimensional settings, and the selected set size is bounded by a polynomial in $N$ even when $p$ grows exponentially.
  • Because LR-FFS and the general framework apply to CRU, MV-SIS, and CAVS as special cases, the same bias-correction machinery can make existing proportion-weighted screening utilities label-shift robust.

Reading between the lines

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

  • A testable extension: the paper's bias correction fixes the focal class share but not the relative mix of the other $R-1$ classes; if that mix differs across clients, each client's 'not $y_r$' contrast estimates a different mixture, so the aggregated target is a weighted blend rather than the pooled utility. Logging per-client class ratios and comparing federated against pooled estimates under asy
  • Switching the aggregation weights from $h_l=\lfloor n_l/2\rfloor$ to the minimum-variance Mann-Whitney weights $\lambda_{l,r}=12|A_r^l||B_r^l|/(n_l+1)$, which the paper derives but does not use in the main text, should lower estimator variance without changing the target; this is a drop-in modification that the analysis already supports.
  • Because the null distribution of each class contrast is centered at $1/2$ under independence, the permutation-based FDR step could be replaced by an exact finite-sample null calculation, making the screening pipeline deterministic and faster; the paper does not explore this shortcut.
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 / 3 minor

Summary. This paper develops a federated feature screening framework for high-dimensional classification under label shift. It proposes a new utility, LR-FFS, defined as the maximum over classes of |E_{Y=y_r}(F_{Y≠y_r}(X_j)) − 1/2|, and estimates it with client-level U-statistics, one-shot aggregation, and a bias correction based on a parameter π* defined through Eq. (5). The paper proves uniform consistency, sure screening, ranking consistency, and FDR control for this estimator, embeds CRU, MV-SIS, and CAVS as special cases of a general framework, and reports extensive simulations plus a TCGA breast-cancer application.

Significance. The framework is elegant and the one-shot communication scheme is practically attractive. If the target-identification issue is resolved, this would be a useful contribution: it unifies several existing screening methods, gives explicit concentration bounds that match centralized rates, demonstrates robustness to heavy tails and outliers, and releases code. The numerical study is extensive and the computational cost claims are plausible. However, the central claim of screening effectiveness 'comparable to centralized processing' is not supported as stated, because under the paper's label-shift assumption the federated estimator's probability limit is not generally the pooled utility.

major comments (4)
  1. The target γ_{j,r} is not identified under S2. On client l, γ_{j,r}^{(l)} = E_{Y=y_r}(F_{Y≠y_r}^{(l)}(X_j)) with F_{Y≠y_r}^{(l)}(x) = Σ_{y≠y_r}(π_l^y/(1−π_l^r)) F_{Y=y}(x); this depends on the relative proportions of the other R−1 classes, which S2 allows to vary arbitrarily across clients. Eq. (5) matches only the weighted average of π_l^r(1−π_l^r), so E(\bar U_{j,r}) = (Σ_l h_l π_l^r(1−π_l^r) γ_{j,r}^{(l)})/(Σ_l h_l), which equals the pooled U_{j,r} only if γ_{j,r}^{(l)} is constant in l. Lemma 20 and Theorem 11 use a single γ_{j,r}; Condition C1 does not rule out, for example, client 1 with classes {1,2} and client 2 with classes {1,3} when P(X|Y=2)≠P(X|Y=3). Proposition 3 itself only covers the case where the ratios among the remaining R−1 classes are fixed, and that assumption is not stated in S2. The authors should either add an explicit common-relative-mixture assumption to the setup or redefine the target as the client-weighted mixture and prove the sure-screening and ranking-consistency properties for that target.
  2. The relevant set A is defined through F(Y|X), but under S2 there is no single conditional distribution of Y given X: P_l(Y|X) differs across clients because P_l(Y) varies while P(X|Y) is common. The paper should define A directly in terms of the common class-conditional distributions P(X|Y), for instance A={j: P(X_j|Y=y) depends on X_j for some y}, or should specify a pooled prior distribution for Y. Without this, the active set in Theorem 11 is ambiguous.
  3. The proof of Theorem 13 conditions on max_j |\barω_j−ω_j| ≥ cN^{-κ} to bound |\hat A|, but the stated bound follows from small estimation error. On the event max_j |\barω_j−ω_j| ≤ cN^{-κ}/2, every selected feature satisfies ω_j ≥ cN^{-κ}/2, so |\hat A| ≤ (c/2)^{-1}N^κ Σ_j ω_j; the complementary event gives the stated probability. As written, the displayed inequality has the wrong direction and the proof needs correction.
  4. The proof of Theorem 14 consists of a sentence referring to Tong et al. after asserting that the estimates have the same efficiency. The conditions Eϕ*_j = 0.5+o(c_n^{-1}) and c_n/p→0 are not verified for the federated permutation estimator of ϕ_j. Since this theorem is stated as a new FDR guarantee, the proof should either verify these conditions under the paper's assumptions or state them as additional assumptions.
minor comments (3)
  1. [Section 2.4, Proposition 6]
  2. [Section 2.3] The identity E_{Y=y_r}(F_{Y=y_r}(X_j)) = 1/2 presumes that X_j is continuous; please state this regularity condition explicitly.
  3. [Figure 1] The text and caption describe the plotted symbols inconsistently (red triangles versus red circles, blue circles versus blue dots); please align them.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the LR-FFS utility is newly defined and its concentration theory is derived from standard inequalities without fitting the claimed conclusion; the label-shift target-identifiability gap is a correctness concern, not a circular step.

full rationale

The central derivation chain defines a new statistic omega_j = max_r |E_{Y=y_r}(F_{Y≠y_r}(X_j)) − 1/2| in Section 2.3, estimates the building block gamma_{j,r} by a one-shot weighted U-statistic in Section 2.4, and then proves concentration of the estimator around this utility through explicit Hoeffding-type bounds (Lemmas 19–21, Proposition 9, Theorem 11). No parameter is fitted to the screening outcome that is later 'predicted': the quantity pi*_r in Eq. (5) is defined solely from the client class proportions, and the displayed identity E(bar-U_{j,r}) = U*_{j,r} is an algebraic consequence of that definition, not a claim that the estimator's limit is the centralized pooled utility by construction. Proposition 2 rewrites existing utilities as special cases of the general framework; that is a mathematical identity, not a circular derivation of the new utility. Citations to Li and Xu (2024) and Li et al. (2020b) provide proof techniques and prior screening algorithms, but the decisive concentration argument for the federated estimator is reproduced in the appendix and does not import the theorem being proved as an input. The genuine weakness flagged in the manuscript and by the skeptic is that under S2 the quantity gamma_{j,r} depends on each client's relative mixture of the other R−1 classes, so the bias-corrected aggregate may converge to a weighted mixture of client-specific targets unless a stronger common-target assumption holds; that is an unstated-assumption or correctness/identifiability issue, not a circular reduction of the type required for a circularity finding. Accordingly, no circular step is established and the score is 0.

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

The central claim rests on the standard label-shift setup, sparsity, and signal-strength conditions. The main additional burden is the unstated requirement that the mixture of the other classes be stable across clients for the federated estimator to target the pooled utility. No new physical entities or ad hoc parameters are introduced.

assumptions (5)
  • domain assumption S1 sparsity: only a few features are relevant, with s = |A| ≪ p.
    Standard in feature screening; stated in Section 2.1 and used throughout.
  • domain assumption S2 label shift with common conditional distribution: P(X|Y) is constant across clients while P(Y) varies.
    Defines the problem setting in Section 2.1.
  • ad hoc to paper Unstated common mixture among the other classes: the relative proportions of the R−1 classes other than y_r are fixed across clients, or the target is the weighted mixture.
    Needed for a single common target utility across clients; not stated in S2 and only partially addressed by Proposition 3. Enters at Eq. (5) and the definition of γ_{j,r}.
  • domain assumption Conditions C1-C4: bounds on class proportions, minimum signal strength, R growth, and utility gap.
    Standard sure screening conditions, listed in Section 3, used in Theorems 11-16.
  • standard math Hoeffding's inequality and Hoeffding's lemma.
    Used in concentration proofs in Lemmas 17-19 and their generalizations.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Label-shift robust federated feature screening for high-dimensional classification." pith.science (2026). https://pith.science/paper/2N4IICM5

@misc{pith2026250600379,
  author       = {Pith},
  title        = {Pith review of: Label-shift robust federated feature screening for high-dimensional classification},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2N4IICM5}},
  note         = {Machine review of arXiv:2506.00379}
}
read the original abstract

Distributed and federated learning are important tools for high-dimensional classification of large datasets. To reduce computational costs and overcome the curse of dimensionality, feature screening plays a pivotal role in eliminating irrelevant features during data preprocessing. However, data heterogeneity, particularly label shifting across different clients, presents significant challenges for feature screening. This paper introduces a general framework that unifies existing screening methods and proposes a novel utility, label-shift robust federated feature screening (LR-FFS), along with its federated estimation procedure. The framework facilitates a uniform analysis of methods and systematically characterizes their behaviors under label shift conditions. Building upon this framework, LR-FFS leverages conditional distribution functions and expectations to address label shift without adding computational burdens and remains robust against model misspecification and outliers. Additionally, the federated procedure ensures computational efficiency and privacy protection while maintaining screening effectiveness comparable to centralized processing. We also provide a false discovery rate (FDR) control method for federated feature screening. Experimental results and theoretical analyses demonstrate LR-FFS's superior performance across diverse client environments, including those with varying class distributions, sample sizes, and missing categorical data.

Figures

Figures reproduced from arXiv: 2506.00379 by the authors.

Figure 1
Figure 1. Impact of label shift on feature screening methods. First row: Relative deviation of utility estimates across heterogeneity levels. Second row: Utility distributions for relevant (red circles) and irrelevant (blue dots) features at selected hetero￾geneity levels. Our paper’s main contributions are threefold. First, we propose a general distributed variable screening framework that unifies existing methods such as CR… view at source ↗
Figure 2
Figure 2. Simulation results for Example 1, left plot displays utility values while right plot shows their time consumption. The horizontal axis indicates the number of seg￾ments. 2.4.1 FDR control In this subsection, we consider how to achieve more precise control over the FDR, referring to Zhu et al. (2011) and Tong et al. (2023) for the introduction of the FDR method. Specifically, for each feature Xj , we independently sh… view at source ↗
Figure 3
Figure 3. Simulation results for Setting (b) in Example 2, proportion of each category follows Dirichlet distribution among different clients. First row represents SSR and second row represents log(wRank). In Setting (a), where outliers are absent, PSIS shows robustness against label shift and achieves effective screening. However, in Settings (b) and (c), where features exhibit heavy￾tailed distributions or outliers, PSIS be… view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Simulation results for Settings (e) and (f) in Example 3, where first row represents SSR and second row represents wRank. As expected, when correlations exist among features, both configurations present greater challenges for accurate feature screening. Due to signific…
Figure 5
Figure 5. Figure 5: Proportions of Subtypes in different institutions (coefficient of contingency: 0.336, p-value of Pearson’s chi-square test: 3.267e − 06), high coefficient of contingency and extremely low p-value both indicate that distribution of subtypes among different hospitals is …
Figure 6
Figure 6. Figure 6: Classification accuracy for different screening methods in TCGA example by KNN [PITH_FULL_IMAGE:figures/full_fig_p026_6.png]
Figure 7
Figure 7. Figure 7: The simulation results for setting (a), where the first row represents SSR and the second row represents wRank. In the presence of noise, the wRank of the PSIS method exceeded 7000. To ensure the clarity of the visualization, the PSIS (Noise) is not shown here. 45 [PI…
Figure 8
Figure 8. Figure 8: The simulation results for setting (f) and (g), where the first row represents SSR and the second row represents log(wRank). Example 5. In this example, we simulate based on setting (a) in example 2 where R = 6, considering simulation results for different weight selec…

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

57 extracted references · 47 canonical work pages

  1. [1]

    write newline

    " write newline "" before.all 'output.state := FUNCTION format.url url empty "" url if FUNCTION article output.bibitem format.authors "author" output.check author format.key output output.year.check new.block format.title "title" output.check new.block crossref missing format.jour.vol output format.article.crossref output.nonnull format.pages output if ne...

  2. [2]

    Federated learning based on dynamic regularization

    Durmus Alp Emre Acar, Yue Zhao, Ramon Matas Navarro, Matthew Mattina, Paul N Whatmough, and Venkatesh Saligrama. Federated learning based on dynamic regularization. arXiv preprint arXiv:2111.04263, 2021

  3. [3]

    Controlling the false discovery rate via knockoffs

    Rina Foygel Barber and Emmanuel J Cand \`e s. Controlling the false discovery rate via knockoffs. The Annals of statistics, 43 0 (5): 0 2055--2085, 2015

  4. [4]

    Federated learning of predictive models from federated electronic health records

    Theodora S Brisimi, Ruidi Chen, Theofanie Mela, Alex Olshevsky, Ioannis Ch Paschalidis, and Wei Shi. Federated learning of predictive models from federated electronic health records. International journal of medical informatics, 112: 0 59--67, 2018

  5. [5]

    Distributed algorithms for u-statistics-based empirical risk minimization

    Lanjue Chen, Alan TK Wan, Shuyi Zhang, and Yong Zhou. Distributed algorithms for u-statistics-based empirical risk minimization. Journal of Machine Learning Research, 24 0 (263): 0 1--43, 2023

  6. [6]

    Distributed statistical inference for massive data

    Song Xi Chen and Liuhua Peng. Distributed statistical inference for massive data. The Annals of Statistics, 49 0 (5): 0 2851--2869, 2021

  7. [7]

    Distributed high-dimensional regression under a quantile loss function

    Xi Chen, Weidong Liu, Xiaojun Mao, and Zhuoyi Yang. Distributed high-dimensional regression under a quantile loss function. Journal of Machine Learning Research, 21 0 (182): 0 1--43, 2020

  8. [8]

    Model-free feature screening for ultrahigh dimensional discriminant analysis

    Hengjian Cui, Runze Li, and Wei Zhong. Model-free feature screening for ultrahigh dimensional discriminant analysis. Journal of the American Statistical Association, 110 0 (510): 0 630--641, 2015

Show all 57 references
  1. [9]

    Imagenet: A large-scale hierarchical image database

    Jia Deng, Wei Dong, Richard Socher, Li-Jia Li, Kai Li, and Li Fei-Fei. Imagenet: A large-scale hierarchical image database. In 2009 IEEE conference on computer vision and pattern recognition, pages 248--255. Ieee, 2009

  2. [10]

    Distributed variable screening for generalized linear models, 2024

    Tianbo Diao, Lianqiang Qu, Bo Li, and Liuquan Sun. Distributed variable screening for generalized linear models, 2024

  3. [11]

    High dimensional classification using features annealed independence rules

    Jianqing Fan and Yingying Fan. High dimensional classification using features annealed independence rules. Annals of statistics, 36 0 (6): 0 2605, 2008

  4. [12]

    Ultrahigh dimensional feature selection: beyond the linear model

    Jianqing Fan, Richard Samworth, and Yichao Wu. Ultrahigh dimensional feature selection: beyond the linear model. The Journal of Machine Learning Research, 10: 0 2013--2038, 2009

  5. [13]

    Nonparametric independence screening in sparse ultra-high-dimensional additive models

    Jianqing Fan, Yang Feng, and Rui Song. Nonparametric independence screening in sparse ultra-high-dimensional additive models. Journal of the American Statistical Association, 106 0 (494): 0 544--557, 2011

  6. [14]

    Mining big data: current status, and forecast to the future

    Wei Fan and Albert Bifet. Mining big data: current status, and forecast to the future. ACM SIGKDD explorations newsletter, 14 0 (2): 0 1--5, 2013

  7. [15]

    Local model poisoning attacks to \ Byzantine-Robust \ federated learning

    Minghong Fang, Xiaoyu Cao, Jinyuan Jia, and Neil Gong. Local model poisoning attacks to \ Byzantine-Robust \ federated learning. In 29th USENIX security symposium (USENIX Security 20), pages 1605--1622, 2020

  8. [16]

    Robustly federated learning model for identifying high-risk patients with postoperative gastric cancer recurrence

    Bao Feng, Jiangfeng Shi, Liebin Huang, Zhiqi Yang, Shi-Ting Feng, Jianpeng Li, Qinxian Chen, Huimin Xue, Xiangguang Chen, Cuixia Wan, et al. Robustly federated learning model for identifying high-risk patients with postoperative gastric cancer recurrence. Nature Communications...

  9. [17]

    Federated learning for medical image analysis: A survey

    Hao Guan, Pew-Thian Yap, Andrea Bozoki, and Mingxia Liu. Federated learning for medical image analysis: A survey. Pattern Recognition, 151: 0 110424, 2024

  10. [18]

    Threshold selection in feature screening for error rate control

    Xu Guo, Haojie Ren, Changliang Zou, and Runze Li. Threshold selection in feature screening for error rate control. Journal of the American Statistical Association, 118 0 (543): 0 1773--1785, 2023

  11. [19]

    Federated visual classification with real-world data distribution

    Tzu-Ming Harry Hsu, Hang Qi, and Matthew Brown. Federated visual classification with real-world data distribution. In Computer Vision--ECCV 2020: 16th European Conference, Glasgow, UK, August 23--28, 2020, Proceedings, Part X 16, pages 76--92. Springer, 2020

  12. [20]

    A distributed one-step estimator

    Cheng Huang and Xiaoming Huo. A distributed one-step estimator. Mathematical Programming, 174: 0 41--76, 2019

  13. [21]

    Learn from others and be yourself in heterogeneous federated learning

    Wenke Huang, Mang Ye, and Bo Du. Learn from others and be yourself in heterogeneous federated learning. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 10143--10153, 2022

  14. [22]

    Communication-efficient distributed statistical inference

    Michael I Jordan, Jason D Lee, and Yun Yang. Communication-efficient distributed statistical inference. Journal of the American Statistical Association, 114 0 (526): 0 668--681, 2019

  15. [23]

    Advances and open problems in federated learning

    Peter Kairouz, H Brendan McMahan, Brendan Avent, Aur \'e lien Bellet, Mehdi Bennis, Arjun Nitin Bhagoji, Kallista Bonawitz, Zachary Charles, Graham Cormode, Rachel Cummings, et al. Advances and open problems in federated learning. Foundations and trends in machine learning, 14...

  16. [24]

    Scaffold: Stochastic controlled averaging for federated learning

    Sai Praneeth Karimireddy, Satyen Kale, Mehryar Mohri, Sashank Reddi, Sebastian Stich, and Ananda Theertha Suresh. Scaffold: Stochastic controlled averaging for federated learning. In International conference on machine learning, pages 5132--5143. PMLR, 2020

  17. [25]

    Federated learning on non-iid data silos: An experimental study

    Qinbin Li, Yiqun Diao, Quan Chen, and Bingsheng He. Federated learning on non-iid data silos: An experimental study. In 2022 IEEE 38th international conference on data engineering (ICDE), pages 965--978. IEEE, 2022

  18. [26]

    Federated optimization in heterogeneous networks

    Tian Li, Anit Kumar Sahu, Manzil Zaheer, Maziar Sanjabi, Ameet Talwalkar, and Virginia Smith. Federated optimization in heterogeneous networks. Proceedings of Machine learning and systems, 2: 0 429--450, 2020 a

  19. [27]

    Privacy-preserving federated brain tumour segmentation

    Wenqi Li, Fausto Milletar \` , Daguang Xu, Nicola Rieke, Jonny Hancox, Wentao Zhu, Maximilian Baust, Yan Cheng, S \'e bastien Ourselin, M Jorge Cardoso, et al. Privacy-preserving federated brain tumour segmentation. In Machine Learning in Medical Imaging: 10th International Wo...

  20. [28]

    Feature screening with conditional rank utility for big-data classification

    Xingxiang Li and Chen Xu. Feature screening with conditional rank utility for big-data classification. Journal of the American Statistical Association, 119 0 (546): 0 1385--1395, 2024

  21. [29]

    Distributed feature screening via componentwise debiasing

    Xingxiang Li, Runze Li, Zhiming Xia, and Chen Xu. Distributed feature screening via componentwise debiasing. Journal of machine learning research, 21 0 (24): 0 1--32, 2020 b

  22. [30]

    A selective overview of feature screening for ultrahigh-dimensional data

    JingYuan Liu, Wei Zhong, and RunZe Li. A selective overview of feature screening for ultrahigh-dimensional data. Science China Mathematics, 58: 0 1--22, 2015

  23. [31]

    Model-free feature screening and fdr control with knockoff features

    Wanjun Liu, Yuan Ke, Jingyuan Liu, and Runze Li. Model-free feature screening and fdr control with knockoff features. Journal of the American Statistical Association, 117 0 (537): 0 428--443, 2022

  24. [32]

    Real-world image datasets for federated learning

    Jiahuan Luo, Xueyang Wu, Yun Luo, Anbu Huang, Yunfeng Huang, Yang Liu, and Qiang Yang. Real-world image datasets for federated learning. arXiv preprint arXiv:1910.11089, 2019

  25. [33]

    No fear of heterogeneity: Classifier calibration for federated learning with non-iid data

    Mi Luo, Fei Chen, Dapeng Hu, Yifan Zhang, Jian Liang, and Jiashi Feng. No fear of heterogeneity: Classifier calibration for federated learning with non-iid data. Advances in Neural Information Processing Systems, 34: 0 5972--5984, 2021

  26. [34]

    The fused kolmogorov filter: A nonparametric model-free screening method

    Qing Mai and Hui Zou. The fused kolmogorov filter: A nonparametric model-free screening method. Annals of Statistics, 43 0 (4): 0 1471--1497, 2015

  27. [35]

    Image-based crop disease detection with federated learning

    Denis Mamba Kabala, Adel Hafiane, Laurent Bobelin, and Rapha \"e l Canals. Image-based crop disease detection with federated learning. Scientific Reports, 13 0 (1): 0 19220, 2023

  28. [36]

    Communication-efficient learning of deep networks from decentralized data

    Brendan McMahan, Eider Moore, Daniel Ramage, Seth Hampson, and Blaise Aguera y Arcas. Communication-efficient learning of deep networks from decentralized data. In Artificial intelligence and statistics, pages 1273--1282. PMLR, 2017

  29. [37]

    Communication-efficient distributed ai strategies for the iot edge

    Christine Mwase, Yi Jin, Tomi Westerlund, Hannu Tenhunen, and Zhuo Zou. Communication-efficient distributed ai strategies for the iot edge. Future Generation Computer Systems, 131: 0 292--308, 2022

  30. [38]

    Lightweight federated learning for stis/hiv prediction

    Thi Phuoc Van Nguyen, Wencheng Yang, Zhaohui Tang, Xiaoyu Xia, Amy B Mullens, Judith A Dean, and Yan Li. Lightweight federated learning for stis/hiv prediction. Scientific Reports, 14 0 (1): 0 6560, 2024

  31. [39]

    Ultrahigh-dimensional multiclass linear discriminant analysis by pairwise sure independence screening

    Rui Pan, Hansheng Wang, and Runze Li. Ultrahigh-dimensional multiclass linear discriminant analysis by pairwise sure independence screening. Journal of the American Statistical Association, 111 0 (513): 0 169--179, 2016

  32. [40]

    Distributed conditional feature screening via pearson partial correlation with fdr control

    Naiwen Pang and Xiaochao Xia. Distributed conditional feature screening via pearson partial correlation with fdr control. arXiv preprint arXiv:2403.05792, 2024

  33. [41]

    A systematic review of federated learning in the healthcare area: From the perspective of data properties and applications

    Prayitno, Chi-Ren Shyu, Karisma Trinanda Putra, Hsing-Chung Chen, Yuan-Yu Tsai, KSM Tozammel Hossain, Wei Jiang, and Zon-Yin Shae. A systematic review of federated learning in the healthcare area: From the perspective of data properties and applications. Applied Sciences, 11 0...

  34. [42]

    The future of digital health with federated learning

    Nicola Rieke, Jonny Hancox, Wenqi Li, Fausto Milletari, Holger R Roth, Shadi Albarqouni, Spyridon Bakas, Mathieu N Galtier, Bennett A Landman, Klaus Maier-Hein, et al. The future of digital health with federated learning. NPJ digital medicine, 3 0 (1): 0 1--7, 2020

  35. [43]

    Identification of anonymous mri research participants with face-recognition software

    Christopher G Schwarz, Walter K Kremers, Terry M Therneau, Richard R Sharp, Jeffrey L Gunter, Prashanthi Vemuri, Arvin Arani, Anthony J Spychalla, Kejal Kantarci, David S Knopman, et al. Identification of anonymous mri research participants with face-recognition software. New ...

  36. [44]

    Federated generalized face presentation attack detection

    Rui Shao, Pramuditha Perera, Pong C Yuen, and Vishal M Patel. Federated generalized face presentation attack detection. IEEE Transactions on Neural Networks and Learning Systems, 35 0 (1): 0 103--116, 2022

  37. [45]

    Cancer statistics, 2024

    Rebecca L Siegel, Angela N Giaquinto, and Ahmedin Jemal. Cancer statistics, 2024. CA: a cancer journal for clinicians, 74 0 (1), 2024

  38. [46]

    Communication-efficient distributed deep learning: A comprehensive survey

    Zhenheng Tang, Shaohuai Shi, Wei Wang, Bo Li, and Xiaowen Chu. Communication-efficient distributed deep learning: A comprehensive survey. arXiv preprint arXiv:2003.06307, 2020

  39. [47]

    Model-free conditional feature screening with fdr control

    Zhaoxue Tong, Zhanrui Cai, Songshan Yang, and Runze Li. Model-free conditional feature screening with fdr control. Journal of the American Statistical Association, 118 0 (544): 0 2575--2587, 2023

  40. [48]

    A survey on distributed machine learning

    Joost Verbraeken, Matthijs Wolting, Jonathan Katzy, Jeroen Kloppenburg, Tim Verbelen, and Jan S Rellermeyer. A survey on distributed machine learning. Acm computing surveys (csur), 53 0 (2): 0 1--33, 2020

  41. [49]

    Distributed and secure federated learning for wireless computing power networks

    Peng Wang, Wen Sun, Haibin Zhang, Wenqiang Ma, and Yan Zhang. Distributed and secure federated learning for wireless computing power networks. IEEE Transactions on Vehicular Technology, 72 0 (7): 0 9381--9393, 2023

  42. [50]

    Category-adaptive variable screening for ultra-high dimensional heterogeneous categorical data

    Jinhan Xie, Yuanyuan Lin, Xiaodong Yan, and Niansheng Tang. Category-adaptive variable screening for ultra-high dimensional heterogeneous categorical data. Journal of the American Statistical Association, 115 0 (530): 0 747--760, 2020

  43. [51]

    Federated learning for healthcare informatics

    Jie Xu, Benjamin S Glicksberg, Chang Su, Peter Walker, Jiang Bian, and Fei Wang. Federated learning for healthcare informatics. Journal of healthcare informatics research, 5: 0 1--19, 2021

  44. [52]

    Byzantine-robust distributed learning: Towards optimal statistical rates

    Dong Yin, Yudong Chen, Ramchandran Kannan, and Peter Bartlett. Byzantine-robust distributed learning: Towards optimal statistical rates. In International Conference on Machine Learning, pages 5650--5659. PMLR, 2018

  45. [53]

    Optimal distributed subsampling for maximum quasi-likelihood estimators with massive data

    Jun Yu, HaiYing Wang, Mingyao Ai, and Huiming Zhang. Optimal distributed subsampling for maximum quasi-likelihood estimators with massive data. Journal of the American Statistical Association, 117 0 (537): 0 265--276, 2022

  46. [54]

    Multiple bayesian discriminant functions for high-dimensional massive data classification

    Jianfei Zhang, Shengrui Wang, Lifei Chen, and Patrick Gallinari. Multiple bayesian discriminant functions for high-dimensional massive data classification. Data mining and knowledge discovery, 31: 0 465--501, 2017

  47. [55]

    Federated learning with label distribution skew via logits calibration

    Jie Zhang, Zhiqi Li, Bo Li, Jianghe Xu, Shuang Wu, Shouhong Ding, and Chao Wu. Federated learning with label distribution skew via logits calibration. In International Conference on Machine Learning, pages 26311--26329. PMLR, 2022

  48. [56]

    Model-free feature screening for ultrahigh-dimensional data

    Li-Ping Zhu, Lexin Li, Runze Li, and Li-Xing Zhu. Model-free feature screening for ultrahigh-dimensional data. Journal of the American Statistical Association, 106 0 (496): 0 1464--1475, 2011

  49. [57]

    Feature screening for massive data analysis by subsampling

    Xuening Zhu, Rui Pan, Shuyuan Wu, and Hansheng Wang. Feature screening for massive data analysis by subsampling. Journal of Business & Economic Statistics, 40 0 (4): 0 1892--1903, 2022

Pith tools

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