PIcsC: Partitioning-Induced Covariate Shift Correction

Behraj Khan; Behroz Mirza; Syed Ahmad Chan Bukhari; Tahir Qasim Syed

REVIEW 4 major objections 5 minor 48 references

PIcsC: Partitioning-Induced Covariate Shift Correction

T0 review · 4 major / 5 minor · reviewed 2026-08-01 · deepseek-v4-flash

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

Pith's one-line read The paper argues that the act of splitting data into batches, cross-validation folds, or federated client shards itself induces a measurable covariate shift, and that a single Fisher-information-based regularizer—estimated from fragment-loc

desk verdict The FIM-KL derivation collapses on inspection; the actual regularizer is EWC with a new trigger, and the experiments tune hyperparameters on the evaluation data. read the letter →

arxiv 2607.25441 v1 pith:LN6P3FAR submitted 2026-07-28 cs.LG

Behraj Khan , Behroz Mirza , Syed Ahmad Chan Bukhari , Tahir Qasim Syed This is my paper

classification cs.LG

keywords distributionshiftcovariateFisherinformationfederatedlearningcross-validationregularizationCramér-Raoboundnon-IIDdata

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper argues that covariate shift can be caused purely by how data is partitioned—whether a central dataset is split into batches or cross-validation folds, or data natively lives on separate federated clients—and that this 'partition-induced' shift is a single phenomenon with a single remedy. That remedy is a regularizer built from the Fisher Information Matrix, estimated locally on each fragment from gradient statistics alone, which penalizes divergence from a running reference. The authors claim that this reduces fragmentation-induced performance degradation by more than 20% (batches) and 25% (folds), and outperforms FedAvg, FedProx, and SCAFFOLD by 3–5 percentage points on seven federated benchmarks without personalization. A conditional variant detects when shift is significant and regularizes only then, keeping memory constant.

What carries the argument

The central object is the diagonal empirical Fisher Information Matrix (FIM)—the averaged squared per-parameter gradients of the log-likelihood over a fragment's local data—used as a tractable stand-in for the KL divergence between a fragment's parameter posterior and a reference (via the Cramér–Rao lower bound). It carries the argument by turning an intractable distributional comparison into a computable penalty that regularizes the loss on each fragment. A conditional extension multiplies the Frobenius change in the FIM by the KL divergence between consecutive fragments into a trigger τt, which decides when to apply a smoothed Cramér–Rao-anchored quadratic regularizer (Eq. 16).

What would settle it

Measure the true KL divergence between empirical covariate distributions of two fragments of a fixed dataset, and compare it with the diagonal-FIM estimate used in Algorithm 1; if the two do not correlate across datasets, the regularizer is not correcting the shift it claims to target. A second check: shuffle the fragment order and repeat—if the gains persist regardless of order, the effect comes from smoothing, not from shift correction.

Watch

Extended reading notes

Core claim

The paper's central claim is that fragmenting data—by design (batches/folds) or by necessity (federated clients)—induces covariate shift between fragments that degrades learning, and that this shift can be quantified and corrected through a single estimator: the diagonal empirical Fisher Information Matrix computed from each fragment's local gradients. Substituting the Cramér–Rao bound for the posterior covariance lets the paper replace the KL divergence between a fragment's parameter distribution and a Gaussian reference with the FIM (Eq. 5), producing a penalized loss (Eq. 7) that accumulates information across fragments. The authors report that this correction restores more than 20% of ba

Load-bearing premise

The method's validity rests on the assumption that a diagonal empirical Fisher matrix computed on one fragment faithfully approximates the KL divergence between that fragment's distribution and a reference—an assumption stated in Section 2.2 with no formula connecting the FIM to the pairwise divergence actually used in the algorithm.

Editorial extensions

If this is right

If the core claim holds, standard cross-validation on fragmented data can recover over 20–25% of the accuracy lost to splitting, merely by adding a Fisher-based penalty.
Federated learning systems could correct statistical heterogeneity without exchanging raw data or per-client personalization, since the estimator uses only local gradient statistics and O(d) communication.
Batch/fold and federated fragmentation become the same optimization problem, allowing methods and insights to transfer between the two literatures.
The conditional trigger provides an on/off switch for regularization, preventing unnecessary drift when fragments are not meaningfully shifted.
The double-mitigation result suggests the same penalty corrects natural covariate shift and partition-induced shift simultaneously.

Reading between the lines

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

The FIM regularizer may in practice function more as a parameter-smoothing term that stabilizes optimization across fragments than as a precise estimator of distributional divergence; an ablation replacing the FIM with a random or constant penalty of matched magnitude would test whether the mechanism is specifically information-theoretic.
If the monofragment estimator truly captures partition divergence, it could be used as a diagnostic: computing per-fragment diagonal FIMs would let practitioners measure shift severity before deciding whether cross-validation or federated aggregation is trustworthy.
The conditional trigger suggests a general recipe—detect shift with a cheap local statistic, then regularize—that could be ported to other divergence measures (e.g., MMD or contrastive losses) when Fisher information is unreliable, such as with transformer-scale models.
Reordering fragments or re-using them in a continual-learning loop would directly probe whether the improvement comes from covariate correction or from a form of experience replay; the paper flags this as future work.

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.

Desk Editor's Note

read the letter

Quick take: the paper's stated mechanism—using Fisher information to estimate KL divergence between data fragments—does not exist in the derivation. What is left is an EWC-style quadratic penalty with a heuristic trigger, and the experimental setup makes the headline numbers look better than they are.

The good stuff: treating batch/fold fragmentation and federated client heterogeneity as instances of the same distribution-shift problem is a framing worth thinking about. The conditional trigger in Eq. (14), a product of FIM change and KL, is a reasonable heuristic for deciding when to regularize. The experimental scope is large—49 datasets, ablations, and a clear statement of limitations in Section 5.7. The authors are transparent about the quadratic cost and the diagonal-FIM approximation.

But the central derivation fails. In Section 2.2, the CRLB is substituted into a Gaussian KL as if an inequality were an equality, and the step from Eq. (5) to Eq. (7) simply drops the divergence and penalizes the FIM itself. There is no formula connecting a fragment's FIM to a pairwise KL between two fragments, so Algorithm 1's d_ij is undefined. The convexity argument in Section 2.3 only shows that the FIM is PSD, which is not the same as the penalty being convex in the parameters—though that is secondary. Worse, Eq. (16) is exactly the EWC penalty, and EWC is not cited. Claiming this as a new Fisher-based shift-correction method without acknowledging EWC is a serious problem.

There is also a load-bearing issue in the experiments. The 36–60% accuracy drops from fragmentation are more plausibly due to training on 5% or 10% of the data than to covariate shift. The paper admits in Section 5.7 that the unfragmented baseline (BL1) is competitive with PIcsC, meaning the method only recovers some of the loss caused by reduced sample support. And λ, α, γ are tuned on the datasets being evaluated, so the reported gains are fitted, not predicted. The federated comparison lacks details on rounds and client sampling.

Who is this for? A reader looking for regularizers in federated learning might mine the EWC-plus-trigger idea, but the paper as a whole does not demonstrate partition-induced covariate shift correction. I would not send this to peer review in its current form; it needs a complete re-derivation and a fairer baseline before it is worth referee time.

Referee Report

4 major / 5 minor

Summary. The paper proposes PIcsC, a Fisher-information-based regularization framework intended to correct 'partition-induced covariate shift' that arises when training data are fragmented into batches, cross-validation folds, or federated client shards. The method approximates a KL divergence between fragment distributions using the Fisher Information Matrix (FIM), adds an FIM-based penalty to the loss, and extends this to a conditional trigger that activates regularization only when a composite FIM/KL signal exceeds a threshold. Experiments are reported on 49 datasets, covering batch, fold, and federated settings, with claims of 20–25% reductions in fragmentation-induced degradation and 3–5 percentage point gains over FedAvg, FedProx, and SCAFFOLD on seven federated benchmarks. The paper also argues that batch/fold and client/node fragmentation are instances of a single phenomenon remediable by one estimator.

Significance. If the central claim were valid, the paper would offer an attractive unified mechanism: a single FIM-based estimator requiring only local gradient statistics, with O(d) memory, applicable to both centralized and distributed fragmentation. The empirical scope is broad (49 datasets, multiple baselines, and federated benchmarks), and the paper clearly states implementation details and limitations. However, the advertised mechanism is not derived: no formula connects the FIM to the pairwise fragment divergence d_ij that Algorithm 1 penalizes, and Eq. (7) is a single-distribution penalty rather than a pairwise divergence. The evaluation also selects key hyperparameters on the same datasets used for the headline claims. As a result, the empirical improvements, even if reproducible, are not evidence for the paper's stated correction mechanism.

major comments (4)

[Section 2.3, Eqs. (8)-(13)] The central estimator is never derived. Eq. (5) replaces V(θhat) in Eq. (4) by I^{-1}(θ) using the CRLB (2), but (2) is an inequality; treating a lower bound as equality is unjustified. More importantly, Eq. (4) is a KL divergence between an arbitrary p(θ) and a Gaussian centered at θ; it is not a divergence between P(D_i) and P(D_j). Eq. (7) then drops the divergence entirely and penalizes the FIM alone. Algorithm 1 line 9 defines d_ij as D_KL(P(D_i)||P(D_j)) and cites Eqs. (1)-(6), but none of those equations depends on a second fragment: the FIM computed on D_i is unchanged if D_j is replaced by any other fragment. Thus the regularizer does not target the advertised pairwise partition divergence. Any empirical gains could arise from a stability regularizer — Eq. (16) is an EWC-style quadratic penalty — rather than from correcting partition-induced covariate shift.
[Section 2.3] The convexity claim is a category error. The derivation shows I(θ) ≥ 0, i.e., that Fisher information is positive semidefinite at a point. This does not show that the penalty term in Eq. (7), viewed as a function of θ, is convex; that would require the Hessian of the penalty to be PSD. The nonnegativity of KL divergence in Eq. (10) is unrelated. Without a correct convexity argument, the optimization-related claims in Section 2.3 are unsupported.
[Section 4.4 / Section 5.1] The claim that random batching induces covariate shift, with drops of 36–60% attributed to that shift, is not supported. Fragmenting an iid dataset changes per-fragment sample size and estimation variance; the declines in Table 1 (e.g., CIFAR-10 from 71.5 to 49.9 when training on 5% of the data) are consistent with training on far fewer examples, not with a change in the covariate distribution P(x). The existence of 'partition-induced covariate shift' is inferred from the very baseline the method is designed to beat, without any direct distributional measurement. This circularity undermines the interpretation of Experiments E1–E4.
[Section 4.5 / Section 5.5] The headline improvements are fitted, not predicted. λ is selected by evaluating PIcsC on the same datasets used for the main results (Figure 1, Section 5.5), and α and γ are calibrated per federated dataset on validation streams (Section 4.5). The ablation in Section 5.6 also tunes on CIFAR-10/CIFAR-100. No held-out protocol is described that would allow the 3–5 point federated gains or the 20–25% fragmentation reductions to be treated as out-of-sample. The results therefore do not establish predictive performance for the proposed method.

minor comments (5)

The method is inconsistently named PIcsC and PIcsR (e.g., Section 6). Please standardize.
[Algorithm 2] The text says 'Compute diagonal I_{t−1}(θ) on F_t (parameters θ_{t−1})'; the indexing of the FIM and the fragment is confusing. It should be I_t(θ) evaluated with the current parameters.
[Eq. (14)] The composite signal τ_t multiplies a Frobenius norm of FIM difference by a KL divergence. These quantities have different units and scales, and no normalization is discussed. It is unclear how γ can be calibrated meaningfully across datasets.
[Tables 5-6] Several entries appear implausible and may be typos: in Table 5, SVHN with 10 folds reports 19.2 for fold 3; in Table 6, Caltech101 PIcsC reports ~95% while st-CV is ~55%, and CIFAR-100 shows large variance under PIcsC. These values need correction or explanation.
[Section 2.2] The sentence citing reference [22] ('monotonicity of entropy and Fisher information') is used to justify replacing the divergence by the FIM alone, but the connection is not explained. Either spell out the argument or remove the citation.

Circularity Check

2 steps flagged · score 6.0 of 10

The claimed FIM-based pairwise fragment divergence d_ij is defined to be the single-fragment FIM penalty, and the conditional regularizer is EWC renamed.

self definitional [Section 2.2, Eq. (7); Algorithm 1, line 9]
"d ij ←D KL(P(D i)∥P(D j)) ▷FIM-based estimate on fragments Di,D j, Eqs. (1)-(6) ... L total ← Ltotal +L(x, y;θ) +λ dij ▷Eq. (7): task loss plus λ-weighted penalty d ij"

Eq. (7) penalizes only the FIM of a single fragment, -λ∫∂²log p(X|θ)/∂θ∂θᵀ dθ, which has no dependence on any second fragment D_j. Algorithm 1 then labels this quantity as D_KL(P(D_i)||P(D_j)), calling it a 'FIM-based estimate on fragments D_i,D_j'. The paper earlier states it 'replac[es] the full divergence term of (5) by the FIM alone,' so the pairwise divergence that PIcsC claims to estimate is not derived; it is defined to be the single-fragment FIM regularizer. Since I(θ) on D_i is unchanged when D_j is replaced by any other fragment, the 'divergence estimate' is a relabeling of the regularizer, not a computed or predicted quantity.
renaming known result [Section 2.4, Eq. (16); Section 4.5]
"Lt(θ) = E_{(x,y)∼P_{F_t}}[ℓ(f_θ(x),y)] + λ(θ−μ_{t−1})^⊤ I^{(t−1)}_{global}(θ−μ_{t−1}) [Cramér-Rao regularizer, Eq. (2)]"

Eq. (16) is exactly the EWC-style quadratic penalty: a Fisher-matrix-weighted distance from the previous parameters. The paper itself concedes this is 'a standard approximation in Fisher-regularized continual learning (e.g. EWC-style methods).' The 'Cramér-Rao regularizer' label comes from Eq. (2), an inequality on estimator covariance, not from any pairwise divergence between fragment distributions. Thus the conditional mechanism's advertised novelty—combining 'FIM shift with KL divergence to detect significant distribution shifts and activat[e] regularization only when necessary'—reduces to a renamed known regularizer; the KL term in the trigger is not actually computed under the stated privacy constraint that clients transmit only diagonal FIMs and parameters.

full rationale

The paper's own derivation never connects the FIM to a pairwise KL divergence between fragment covariate distributions. Section 2.2 defines I(θ) for a single model at a single parameter value, substitutes the CRLB inequality into a Gaussian-posterior KL as though it were equality, and then, citing an entropy/Fisher monotonicity result, simply 'replac[es] the full divergence term of (5) by the FIM alone.' That replacement is an assumption, not a theorem. Algorithm 1 then labels the resulting single-fragment FIM penalty as d_ij = D_KL(P(D_i)||P(D_j)), so the central 'estimator of partition divergence' is defined to be the regularizer rather than derived from any two-fragment divergence. The conditional variant is likewise an EWC-style quadratic penalty with a FIM-shift trigger; the advertised KL component is not computed under the privacy constraint. These are definitional reductions of the paper's mechanistic claim, not merely missing rigor. No load-bearing self-citation or imported uniqueness theorem is present, and the empirical benchmarks are external, so the method may still act as a useful EWC-style stability regularizer; that independent empirical content prevents a higher score. Hyperparameter selection of λ, α, and γ on validation streams is standard model selection and is not by itself a circularity. Overall, the central 'predictions' about correcting pairwise partition-induced shift reduce by construction, yielding partial circularity.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The method relies on a chain of approximations: CRLB as equality, diagonal empirical Fisher as the FIM, and the FIM as a proxy for data-distribution divergence. None of these is quantified or validated independently. λ, α, γ are tuned on the evaluation data.

free parameters (3)

λ (penalty strength) = 0.1
Calibrated in Figure 1 on the fold-wise KEEL settings and used for all experiments; not swept per architecture or loss (Section 4.5 admits this).
α (FIM smoothing rate) = not reported
Selected on held-out validation stream per federated dataset (Section 4.5); not reported in the paper.
γ (detection threshold) = not reported
Calibrated per federated dataset (Section 4.5); no value given.

assumptions (4)

domain assumption Cramér-Rao bound is used as an equality: V(θ̂) = I^{-1}(θ) when replacing the posterior covariance in the Gaussian KL.
Section 2.2, Eq. (5) substitutes V = I^{-1} into the Gaussian, turning an inequality into an equality; the paper acknowledges this is an approximation in Section 5.7.
domain assumption The diagonal empirical Fisher (squared per-parameter gradients) approximates the true FIM and serves as a proxy for KL divergence between fragments.
Section 4.5 (diagonal empirical Fisher) and Section 2.2 ('replacing the full divergence term by the FIM alone'); the link between the FIM and the pairwise data-divergence dij is asserted, not derived.
ad hoc to paper Randomly fragmenting a dataset into batches/folds induces a meaningful covariate shift distinct from finite-sample effects.
Section 1 and E1 (Section 4.4); the magnitude of the reported drop (36-60%) suggests small-sample degradation; the paper does not separate the two effects.
domain assumption The FIM penalty is convex and preserves optimization properties.
Section 2.3 'proves' convexity by showing I(θ) ≥ 0 (Eq. 13), which is PSD, not convexity of the loss; the argument is a category error but the claim is load-bearing for the 'does not alter monotonicity' statement.

how reviews work

0 comments

Cite this review

Pith. "Pith review of PIcsC: Partitioning-Induced Covariate Shift Correction." pith.science (2026). https://pith.science/paper/LN6P3FAR

@misc{pith2026260725441,
  author       = {Pith},
  title        = {Pith review of: PIcsC: Partitioning-Induced Covariate Shift Correction},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/LN6P3FAR}},
  note         = {Machine review of arXiv:2607.25441}
}

read the original abstract

Covariate shift across training-data partitions biases model selection and parameter estimation in cross-validation, lifelong learning, and federated learning. We propose \textit{Partition-Induced Covariate-shift Correction} (\texttt{PIcsC}), a Fisher information-based regularization framework that mitigates distribution mismatch between data partitions and a reference distribution. \texttt{PIcsC} approximates partition divergence using the Fisher Information Matrix (FIM) and incorporates the resulting statistic as a regularizer during optimization. The same formulation applies to both centrally partitioned datasets (batches or cross-validation folds) and inherently distributed data (federated clients or decentralized nodes), requiring only partition-local gradient statistics rather than raw data. We further introduce a conditional adaptation mechanism that combines FIM shift with KL divergence to detect significant distribution shifts and activates regularization only when necessary. Experiments on more than 40 datasets demonstrate consistent improvements under both natural and synthetic covariate shift. On fragmented batch and fold settings, \texttt{PIcsC} reduces fragmentation-induced performance degradation by more than 20\% and 25\%, respectively. On seven federated learning benchmarks, it consistently outperforms FedAvg, FedProx, and SCAFFOLD by 3 -5 percentage points without requiring client-specific personalization. These results demonstrate that Fisher information provides an effective and unified mechanism for mitigating partition-induced covariate shift across both centralized and distributed learning.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

48 extracted references · 15 linked inside Pith

[1]

Mit Press, ??? (2008)

Qui˜ nonero-Candela, J., Sugiyama, M., Schwaighofer, A., Lawrence, N.D.: Dataset Shift in Machine Learning. Mit Press, ??? (2008)

2008
[2]

In: Algorithmic Learning Theory: 19th International Conference, ALT 2008, Budapest, Hungary, October 13-16, 2008

Cortes, C., Mohri, M., Riley, M., Rostamizadeh, A.: Sample selection bias cor- rection theory. In: Algorithmic Learning Theory: 19th International Conference, ALT 2008, Budapest, Hungary, October 13-16, 2008. Proceedings 19, pp. 38–53 (2008). Springer

2008
[3]

Journal of Machine Learning Research8(5) (2007)

Sugiyama, M., Krauledat, M., M¨ uller, K.-R.: Covariate shift adaptation by impor- tance weighted cross validation. Journal of Machine Learning Research8(5) (2007)

2007
[4]

In: Proceedings of the 24th International Conference on Machine Learning, pp

Bickel, S., Br¨ uckner, M., Scheffer, T.: Discriminative learning for differing training and test distributions. In: Proceedings of the 24th International Conference on Machine Learning, pp. 81–88 (2007)

2007
[5]

Artificial Intelligence in Medicine29(1-2), 81–106 (2003)

Kukar, M.: Transductive reliability estimation for medical diagnosis. Artificial Intelligence in Medicine29(1-2), 81–106 (2003)

2003
[6]

Interfaces48(5), 449–466 (2018)

Rudin, C., Ustun, B.: Optimized scoring systems: Toward trust in machine learning for healthcare and criminal justice. Interfaces48(5), 449–466 (2018)

2018
[7]

arXiv:2011.01961 (2020)

Wong, A., Hryniowski, A., Wang, X.Y.: Insights into fairness through trust: multi-scale trust quantification for financial deep learning. arXiv:2011.01961 (2020)

arXiv 2011
[8]

In: Proceedings of the AAAI Conference on Artificial Intelligence, vol

Rezaei, A., Liu, A., Memarrast, O., Ziebart, B.D.: Robust fairness under covariate shift. In: Proceedings of the AAAI Conference on Artificial Intelligence, vol. 35, pp. 9419–9427 (2021)

2021

Show all 48 references

[9]

Advances in neural information processing systems20(2007) 27

Sugiyama, M., Nakajima, S., Kashima, H., Buenau, P., Kawanabe, M.: Direct importance estimation with model selection and its application to covariate shift adaptation. Advances in neural information processing systems20(2007) 27

2007
[10]

IEEE Transactions on Neural Networks and Learning Systems23(8), 1304–1312 (2012)

Moreno-Torres, J.G., S´ aez, J.A., Herrera, F.: Study on the impact of partition- induced dataset shift on k-fold cross-validation. IEEE Transactions on Neural Networks and Learning Systems23(8), 1304–1312 (2012)

2012
[11]

Advances in Neural Information Processing Systems34, 23664–23678 (2021)

Zhang, M., Marklund, H., Dhawan, N., Gupta, A., Levine, S., Finn, C.: Adap- tive risk minimization: Learning to adapt to domain shift. Advances in Neural Information Processing Systems34, 23664–23678 (2021)

2021
[12]

IEEE transactions on pattern analysis and machine intelligence 44(7), 3366–3385 (2021)

De Lange, M., Aljundi, R., Masana, M., Parisot, S., Jia, X., Leonardis, A., Slabaugh, G., Tuytelaars, T.: A continual learning survey: Defying forgetting in classification tasks. IEEE transactions on pattern analysis and machine intelligence 44(7), 3366–3385 (2021)

2021
[13]

In: Artificial Intelligence and Statistics, pp

McMahan, B., Moore, E., Ramage, D., Hampson, S., Arcas, B.A.: Communication- efficient learning of deep networks from decentralized data. In: Artificial Intelligence and Statistics, pp. 1273–1282 (2017). Pmlr

2017
[14]

arXiv:2205.11101 (2022)

Gupta, S., Ahuja, K., Havaei, M., Chatterjee, N., Bengio, Y.: Fl games: A federated learning framework for distribution shifts. arXiv:2205.11101 (2022)

2022 arXiv
[15]

arXiv preprint arXiv:2411.12377 (2024)

Solans, D., Heikkila, M., Vitaletti, A., Kourtellis, N., Anagnostopoulos, A., Chatzi- giannakis, I., et al.: Non-iid data in federated learning: A systematic review with taxonomy, metrics, methods, frameworks and future directions. arXiv preprint arXiv:2411.12377 (2024)

2024 arXiv
[16]

ACM Computing Surveys56(3), 1–44 (2023)

Ye, M., Fang, X., Du, B., Yuen, P.C., Tao, D.: Heterogeneous federated learning: State-of-the-art and research challenges. ACM Computing Surveys56(3), 1–44 (2023)

2023
[17]

In: International Conference on Machine Learning, pp

Karimireddy, S.P., Kale, S., Mohri, M., Reddi, S., Stich, S., Suresh, A.T.: Scaffold: Stochastic controlled averaging for federated learning. In: International Conference on Machine Learning, pp. 5132–5143 (2020). PMLR

2020
[18]

Proceedings of Machine learning and systems2, 429–450 (2020)

Li, T., Sahu, A.K., Zaheer, M., Sanjabi, M., Talwalkar, A., Smith, V.: Federated optimization in heterogeneous networks. Proceedings of Machine learning and systems2, 429–450 (2020)

2020
[19]

arXiv:1301.3584 (2013)

Pascanu, R., Bengio, Y.: Revisiting natural gradient for deep networks. arXiv:1301.3584 (2013)

2013 arXiv
[20]

arXiv:2107.10578 (2021)

Bukaew, W., Yoo-Kong, S.: One-parameter generalised fisher information. arXiv:2107.10578 (2021)

2021 arXiv
[21]

arXiv:1907.00288 (2019)

Nishiyama, T.: A new lower bound for kullback-leibler divergence based on hammersley-chapman-robbins bound. arXiv:1907.00288 (2019)

1907 arXiv
[22]

arXiv:1610.04174 (2016)

Courtade, T.A.: Monotonicity of entropy and fisher information: a quick proof via 28 maximal correlation. arXiv:1610.04174 (2016)

2016 arXiv
[23]

Cambridge University Press, ??? (2012)

Sugiyama, M., Suzuki, T., Kanamori, T.: Density Ratio Estimation in Machine Learning. Cambridge University Press, ??? (2012)

2012
[24]

Advances in neural information processing systems19(2006)

Huang, J., Gretton, A., Borgwardt, K., Sch¨ olkopf, B., Smola, A.: Correcting sample selection bias by unlabeled data. Advances in neural information processing systems19(2006)

2006
[25]

Journal of statistical planning and inference90(2), 227–244 (2000)

Shimodaira, H.: Improving predictive inference under covariate shift by weighting the log-likelihood function. Journal of statistical planning and inference90(2), 227–244 (2000)

2000
[26]

Advances in neural information processing systems33, 11996–12007 (2020)

Fang, T., Lu, N., Niu, G., Sugiyama, M.: Rethinking importance weighting for deep learning under distribution shift. Advances in neural information processing systems33, 11996–12007 (2020)

2020
[27]

Journal of Information Processing17, 138–155 (2009)

Tsuboi, Y., Kashima, H., Hido, S., Bickel, S., Sugiyama, M.: Direct density ratio estimation for large-scale covariate shift adaptation. Journal of Information Processing17, 138–155 (2009)

2009
[28]

In: The 22nd International Conference on Artificial Intelligence and Statistics, pp

Stojanov, P., Gong, M., Carbonell, J., Zhang, K.: Low-dimensional density ratio estimation for covariate shift correction. In: The 22nd International Conference on Artificial Intelligence and Statistics, pp. 3449–3458 (2019). PMLR

2019
[29]

arXiv:2302.02552 (2023)

Zhang, Y.-J., Zhang, Z.-Y., Zhao, P., Sugiyama, M.: Adapting to continuous covariate shift via online density ratio estimation. arXiv:2302.02552 (2023)

2023 arXiv
[30]

MIT press, ??? (2012)

Sugiyama, M., Kawanabe, M.: Machine Learning in Non-stationary Environments: Introduction to Covariate Shift Adaptation. MIT press, ??? (2012)

2012
[31]

Dataset shift in machine learning 3(4), 5 (2009)

Gretton, A., Smola, A., Huang, J., Schmittfull, M., Borgwardt, K., Sch¨ olkopf, B., et al.: Covariate shift by kernel mean matching. Dataset shift in machine learning 3(4), 5 (2009)

2009
[32]

Neural Networks165, 248–273 (2023)

Faber, K., Corizzo, R., Sniezynski, B., Japkowicz, N.: Vlad: Task-agnostic vae-based lifelong anomaly detection. Neural Networks165, 248–273 (2023)

2023
[33]

arXiv preprint arXiv:2210.04865 (2022)

Basterrech, S., Wo´ zniak, M.: Tracking changes using kullback-leibler divergence for the continual learning. arXiv preprint arXiv:2210.04865 (2022)

2022 arXiv
[34]

Machine Learning111(7), 2561–2599 (2022)

Cano, A., Krawczyk, B.: Rose: robust online self-adjusting ensemble for continual learning on imbalanced drifting data streams. Machine Learning111(7), 2561–2599 (2022)

2022
[35]

http://yann

LeCun, Y.: The mnist database of handwritten digits. http://yann. lecun. com/exdb/mnist/ (1998) 29

1998
[36]

arXiv:1708.07747 (2017)

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

2017 arXiv
[37]

Krizhevsky, A., Hinton, G., et al.: Learning multiple layers of features from tiny images (2009)

2009
[38]

arXiv:1812.01718 (2018)

Clanuwat, T., Bober-Irizar, M., Kitamoto, A., Lamb, A., Yamamoto, K., Ha, D.: Deep learning for classical japanese literature. arXiv:1812.01718 (2018)

2018 arXiv
[39]

Netzer, Y., Wang, T., Coates, A., Bissacco, A., Wu, B., Ng, A.Y.: Reading digits in natural images with unsupervised feature learning (2011)

2011
[40]

CVPR Workshop (2004)

Fei-Fei, L., Fergus, R., Perona, P.: Learning generative visual models from few train- ing examples: An incremental bayesian approach tested on 101 object categories. CVPR Workshop (2004)

2004
[41]

In: Proceedings of the Fourteenth International Conference on Artificial Intelligence and Statistics, pp

Coates, A., Ng, A., Lee, H.: An analysis of single-layer networks in unsupervised feature learning. In: Proceedings of the Fourteenth International Conference on Artificial Intelligence and Statistics, pp. 215–223 (2011). JMLR Workshop and Conference Proceedings

2011
[42]

arXiv:1312.6211 (2013)

Goodfellow, I.J., Mirza, M., Xiao, D., Courville, A., Bengio, Y.: An empiri- cal investigation of catastrophic forgetting in gradient-based neural networks. arXiv:1312.6211 (2013)

2013 arXiv
[43]

arXiv:1906.02337 (2019)

Mu, N., Gilmer, J.: Mnist-c: A robustness benchmark for computer vision. arXiv:1906.02337 (2019)

1906 arXiv
[44]

Proceedings of the International Conference on Learning Representations (2019)

Hendrycks, D., Dietterich, T.: Benchmarking neural network robustness to common corruptions and perturbations. Proceedings of the International Conference on Learning Representations (2019)

2019
[45]

Journal of Multiple-Valued Logic & Soft Computing17(2011)

Alcal´ a-Fdez, J., Fern´ andez, A., Luengo, J., Derrac, J., Garc ´ ıa, S., S´ anchez, L., Herrera, F.: Keel data-mining software tool: data set repository, integration of algorithms and experimental analysis framework. Journal of Multiple-Valued Logic & Soft Computing17(2011)

2011
[46]

arXiv preprint arXiv:1812.01097 (2018)

Caldas, S., Duddu, S.M.K., Wu, P., Li, T., Koneˇ cn` y, J., McMahan, H.B., Smith, V., Talwalkar, A.: Leaf: A benchmark for federated settings. arXiv preprint arXiv:1812.01097 (2018)

2018 arXiv
[47]

arXiv preprint arXiv:2403.03952 (2024)

Hou, Y., Li, J., He, Z., Yan, A., Chen, X., McAuley, J.: Bridging language and items for retrieval and recommendation. arXiv preprint arXiv:2403.03952 (2024)

2024 arXiv
[48]

Advances in Neural Information Processing Systems32(2019) 30

Rabanser, S., G¨ unnemann, S., Lipton, Z.: Failing loudly: An empirical study of methods for detecting dataset shift. Advances in Neural Information Processing Systems32(2019) 30

2019

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.

[1] [1]

Mit Press, ??? (2008)

Qui˜ nonero-Candela, J., Sugiyama, M., Schwaighofer, A., Lawrence, N.D.: Dataset Shift in Machine Learning. Mit Press, ??? (2008)

2008

[2] [2]

In: Algorithmic Learning Theory: 19th International Conference, ALT 2008, Budapest, Hungary, October 13-16, 2008

Cortes, C., Mohri, M., Riley, M., Rostamizadeh, A.: Sample selection bias cor- rection theory. In: Algorithmic Learning Theory: 19th International Conference, ALT 2008, Budapest, Hungary, October 13-16, 2008. Proceedings 19, pp. 38–53 (2008). Springer

2008

[3] [3]

Journal of Machine Learning Research8(5) (2007)

Sugiyama, M., Krauledat, M., M¨ uller, K.-R.: Covariate shift adaptation by impor- tance weighted cross validation. Journal of Machine Learning Research8(5) (2007)

2007

[4] [4]

In: Proceedings of the 24th International Conference on Machine Learning, pp

Bickel, S., Br¨ uckner, M., Scheffer, T.: Discriminative learning for differing training and test distributions. In: Proceedings of the 24th International Conference on Machine Learning, pp. 81–88 (2007)

2007

[5] [5]

Artificial Intelligence in Medicine29(1-2), 81–106 (2003)

Kukar, M.: Transductive reliability estimation for medical diagnosis. Artificial Intelligence in Medicine29(1-2), 81–106 (2003)

2003

[6] [6]

Interfaces48(5), 449–466 (2018)

Rudin, C., Ustun, B.: Optimized scoring systems: Toward trust in machine learning for healthcare and criminal justice. Interfaces48(5), 449–466 (2018)

2018

[7] [7]

arXiv:2011.01961 (2020)

Wong, A., Hryniowski, A., Wang, X.Y.: Insights into fairness through trust: multi-scale trust quantification for financial deep learning. arXiv:2011.01961 (2020)

arXiv 2011

[8] [8]

In: Proceedings of the AAAI Conference on Artificial Intelligence, vol

Rezaei, A., Liu, A., Memarrast, O., Ziebart, B.D.: Robust fairness under covariate shift. In: Proceedings of the AAAI Conference on Artificial Intelligence, vol. 35, pp. 9419–9427 (2021)

2021