Pith. sign in

REVIEW 3 minor 73 references

Defensive Boosting for Online Probabilistic Forecasting

T0 review · 0 major / 3 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read One algorithm unifies online gradient boosting and weak-to-strong boosting

desk verdict A clean one-oracle online boosting algorithm that unifies two previously separate guarantees; the main soft spot is an abstract that overstates the oracle assumption. read the letter →

arxiv 2608.13554 v1 pith:FNWIGLP3 submitted 2026-08-13 cs.LG cs.CCcs.DSstat.ML

Georgy Noarov , Aaron Roth This is my paper · ORCID
classification cs.LGcs.CCcs.DSstat.ML MSC 68T0568Q32
keywords onlineboostingprobabilisticforecastingdefensivemultiaccuracyself-orthogonalityweak-to-stronggradientBrierscore
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 gives a single online algorithm, the Defensive Booster, that produces probability forecasts for binary outcomes chosen by an adaptive adversary and simultaneously obtains the two guarantees that previously required separate boosting methods. On every sequence, its average Brier score competes with the best predictor in the norm-bounded span of the weak class at the same rate as online gradient boosting. Whenever the realized transcript satisfies the smooth weak-learning condition, both its Brier score and its randomized classification error are driven toward zero at the same 1/($gamma^{2}$ epsilon) rate that online classification boosting achieves. This matters because the two guarantees are incomparable; a unified forecaster removes the need to choose a method before seeing the data.

What carries the argument

The load-bearing object is the second-order weak-class oracle: after seeing the current context but before seeing the coefficient, it outputs a weak prediction and guarantees regret against every h in H scaling as a_H $\sqrt$(sum $c_t^{2}$) + b_H. The Defensive Booster wraps this oracle with two scalar adaptive-gradient auditors and a one-dimensional root rule, solving an affine equation in the signed forecast mu so that the aggregated auditor gain is nonpositive no matter what label arrives. The two auditors enforce multiaccuracy and self-orthogonality with errors A_H $\sqrt$(S_T) + B_H and A_S $\sqrt$(S_T) + B_S, where S_T = 4 sum (Y_t - p_t)^2. The identity w_t sigma_t = r_t/2 then converts multiaccuracy into a bound on the edge of the mistake weighting, and the smooth weak-learning condition turns the low-edge witness into a strong-learning guarantee by contrapositive.

What would settle it

Run the Defensive Booster on the paper's binary-aggregation construction, which guarantees edge at least 0.16, and measure the normalized edge of its mistake weighting under the weak class; if the edge exceeds the Theorem 4.4 bound A_H $\sqrt$(T B_T)/(T rho_w), or if Brier and randomized errors fail to shrink as O(1/($gamma^{2}$ T)) under the smooth weak-learning condition, the central claim is false.

Watch

Extended reading notes

Core claim

The Defensive Booster's central discovery is that the dual view of boosting can be made operational online: instead of building an ensemble of weak learners, the forecaster chooses its probability p_t so that two sustained correlations vanish - no weak hypothesis correlates with the signed forecast residuals, and neither does the forecast itself. Multiaccuracy alone turns the residual weights w_t = |Y_t - p_t| into a reweighting of the transcript on which the entire weak class has small edge, so if the algorithm's randomized classification error stays high long enough, those weights form a smooth hard-core witness that the smooth weak-learning condition fails. Adding the self-auditor gives the unconditional span-regret guarantee, because multiaccuracy plus self-orthogonality are exactly the first-order optimality conditions for squared loss. The two guarantees follow from the same two inequalities in Theorem 3.3, not from one another.

Load-bearing premise

The whole construction depends on having a weak-class learner whose regret to every hypothesis in the class scales with the square root of how strongly it was used, not just with the number of rounds; with only a first-order learner, the claimed fast weak-to-strong rate collapses.

Editorial extensions

If this is right

  • Every adaptive sequence gets the gradient-boosting span guarantee: Brier score is at most the best span comparator loss plus O((Lambda A_H + A_S)/sqrt(T)), with a second-order refinement that becomes O(1/T) in the realizable case.
  • Under the (rho, gamma)-smooth weak-learning condition, both Brier score and randomized classification error are at most max{rho, O(A_H^2/(gamma^2 T)), O(B_H/(gamma T))}, so epsilon accuracy needs T = O(1/(gamma^2 epsilon)), the same dependence previously shown optimal for online weak-to-strong boosting.
  • When the forecaster's error remains large, its mistake weights form an ex-post smooth low-edge reweighting of the transcript, certifying that the smooth weak-learning condition cannot hold on that transcript.
  • The strongly adaptive variant gives both guarantees on every contiguous interval up to polylogarithmic factors, and its interval mistake weights localize the hard-core witness to identify where and when weak learnability fails.
  • The algorithm uses one weak-class oracle call plus O(1) arithmetic per round, and in experiments tracks or beats the stronger of the ensemble baselines while being 20-66 times faster per round.

Reading between the lines

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

  • Because the hard-core direction needs only multiaccuracy, any online learner that maintains multiaccuracy could in principle be boosted by the same dual argument; a testable extension would be to replace the Brier root rule with another proper scoring rule.
  • The interval hard-core witnesses double as a change-point detector: on a data stream, a persistent smooth low-edge weighting along a trailing interval indicates that the weak class has stopped being informative, without any separate drift-detection subroutine.
  • The second-order oracle assumption is doing real work, since a first-order oracle would degrade the sample complexity to 1/(gamma^2 epsilon^2); this suggests that online boosting theory may benefit from building data-dependent-regret oracles explicitly rather than treating O(sqrt(T)) online learners as generic black boxes.
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

0 major / 3 minor

Summary. The paper introduces the Defensive Booster, an online probabilistic forecaster for binary (and, by extension, bounded real-valued) outcomes under an adaptive adversary. The algorithm is a black-box reduction from a second-order weak-class oracle (Definition 2.3) together with two scalar adaptive-OGD states. It uses a one-dimensional root rule to enforce two statistical conditions: multiaccuracy with respect to the weak class and self-orthogonality of the forecasts (Theorem 3.3). From these conditions the paper derives an unconditional Brier-score regret bound against the norm-bounded span of the weak class (Theorem 4.1), a hard-core mistake-weighting certificate (Theorem 4.4), and, under the smooth weak-learning condition, a weak-to-strong bound on Brier score and randomized classification error of the form max{rho_0, O(A_H^2/(gamma_0^2 T)), O(B_H/(gamma_0 T))} (Corollary 4.5). It also gives a strongly adaptive interval version (Section 5), separation examples showing the two guarantees are incomparable (Appendix B), and an extensive empirical evaluation on synthetic streams, real binary streams, and chronological regression data, with a released implementation. The claimed contribution is a single one-oracle algorithm that simultaneously obtains online gradient boosting's span guarantee and online weak-to-strong boosting's classification guarantee, at rates matching prior specialized methods.

Significance. If the results hold, the paper is a substantial contribution to online learning and boosting. It unifies two previously separate families of online boosting guarantees in a single simple algorithm, provides the first online boosting theorem derived through the defensive-forecasting / multiaccuracy dual view, and matches the optimal 1/(gamma^2 epsilon) weak-to-strong rate under a clean, explicitly stated oracle model. The proofs are largely self-contained and the main inequalities are checkable: the root sign property, the second-order scalar regret lemma, the multiaccuracy-to-hard-core conversion, and the contrapositive rate argument are all coherent. The paper is also unusually complete on the empirical side: it reports synthetic streams engineered for each guarantee, real-data binary streams, regression extensions, a controlled drift benchmark, runtime comparisons, and code with reproduction commands.

minor comments (3)
  1. [Abstract and Section 1.1] The abstract's opening statement, 'Given an online learning algorithm for a weak hypothesis class H,' is broader than the formal primitive in Definition 2.3, which requires a second-order no-regret oracle whose regret scales as a_H sqrt(sum_t c_t^2) + b_H. An ordinary first-order O(sqrt(T)) learner would only yield the degraded 1/(gamma_0^2 epsilon^2) rate, as the paper itself notes in Section 2. Please qualify the abstract and the informal theorem statements to say 'given a second-order no-regret online learning algorithm,' so that the advertised scope matches the theorem.
  2. [Appendix A.3 / Proposition 5.1] The proof of Proposition 5.1 invokes 'the standard guarantee' of a second-order confidence-rated experts algorithm and cites Gaillard et al. (2014) without stating the exact regret bound or the constant C_0. Since the logarithmic factors in Corollary 5.3 depend on this bound, please state the precise inequality being used or provide a short derivation within the appendix.
  3. [Section 4.3, after Corollary 4.5] The discussion says that the lower bound of Beygelzimer et al. (2015b) shows the 1/(gamma_0^2 epsilon) dependence is unavoidable, and the text does say 'in their model.' Because Definition 2.3 is a stronger oracle than their weak-online-learning model, the lower bound does not apply directly to the present setting; a sentence clarifying that the match is only in the gamma, epsilon rate, not in the exact model, would prevent over-reading.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the weak-to-strong guarantee is a genuine contrapositive of the multiaccuracy hard-core certificate, not a restatement of the oracle assumption.

full rationale

The derivation chain is self-contained. Algorithm 1 is a defensive forecaster whose certificate (Theorem 3.3) is proved from the second-order weak-class oracle (Definition 2.3) and scalar adaptive OGD (Lemma 2.5), using only the root-sign property (Lemma 3.2). The Brier/span guarantee (Theorem 4.1) is obtained by substituting the multiaccuracy and self-orthogonality bounds into the signed convexity inequality for squared loss; no fitted comparator or assumed outcome is reused as a conclusion. The hard-core mistake weighting (Theorem 4.4) follows from the identity w_t sigma_t = r_t/2, converting the multiaccuracy bound into an edge bound on the algorithm's own mistake weights. Corollary 4.5 is exactly the contrapositive: if every rho0-smooth reweighting has edge at least gamma0, then the algorithm's weighting cannot be simultaneously smooth and low-edge, and solving the two inequalities gives the max{rho0, O(1/(gamma0^2 T))} rate. The paper explicitly states that its second-order oracle is stronger than the prior weak-online-learning model used in the lower bound it matches ('although our oracle model is stronger', Section 2), so the rate comparison is a conditional claim about a stated primitive rather than an equation-level identity. Self-citations (Noarov--Roth multicalibration work, Kearns et al. self-orthogonality, Collina et al. lower bounds) are used as context or are accompanied by independent proofs and external citations; none carries the load of a main theorem by itself. I find no circular step.

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

The Defensive Booster introduces no fitted parameters or invented entities. It relies on the specified weak-class oracle and standard online-learning regret bounds; the smooth weak-learning condition is an explicit condition in the theorem, not a hidden axiom.

assumptions (5)
  • domain assumption Adversarial online protocol with arbitrary adaptive sequences.
    Section 2: the environment reveals x_t, the learner predicts p_t, then the outcome is revealed; no distributional assumptions. All guarantees hold pathwise.
  • domain assumption Second-order weak-class oracle (Definition 2.3).
    The central primitive: for coefficients c_t in [-2,2], regret to every h in H is bounded by a_H sqrt(sum c_t^2) + b_H. Instantiations include finite experts and RKHS balls; without this, the weak-to-strong rate degrades.
  • domain assumption Symmetric weak class H.
    Used in Theorem 3.3 for the absolute-value multiaccuracy bound and in Definition 4.3 for edge; the paper notes one can replace H by H union (-H).
  • standard math Scalar adaptive OGD second-order regret (Lemma 2.5).
    Proved in Appendix A.1; used in Theorem 3.3 and in the interval wrapper.
  • standard math Existing second-order confidence-rated experts bounds (Gaillard et al. 2014).
    Proposition 5.1 invokes this prior result for the strongly adaptive wrapper; not self-contained in the paper.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Defensive Boosting for Online Probabilistic Forecasting." pith.science (2026). https://pith.science/paper/FNWIGLP3

@misc{pith2026260813554,
  author       = {Pith},
  title        = {Pith review of: Defensive Boosting for Online Probabilistic Forecasting},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FNWIGLP3}},
  note         = {Machine review of arXiv:2608.13554}
}
abstract

We study online probabilistic forecasting of binary outcomes chosen by an adaptive adversary. Given an online learning algorithm for a weak hypothesis class $H$, we would like to efficiently obtain two incomparable guarantees that existing online boosting techniques provide separately. Online gradient boosting competes in Brier score with the best predictor induced by the span of $H$ on every sequence, but promises nothing when the span does not contain an accurate predictor. Online weak-to-strong boosting drives classification error to zero under a weak-learning condition, but promises little when that condition fails. We give a simple defensive forecasting algorithm, the Defensive Booster, that obtains both guarantees. On every adaptive sequence, its Brier score is competitive with the best prediction induced by the span of $H$ at the same rate as online gradient boosting; simultaneously, whenever the realized transcript satisfies the smooth weak-learning condition, its Brier score and randomized classification error satisfy the same rate guarantee as online classification boosting. This is achieved by operationalizing the "dual view" of boosting: When the algorithm's randomized classification error is persistently high, its mistake weights form a smooth reweighting on which every weak hypothesis has low edge, yielding an ex-post hard-core certificate that the weak-learning condition fails. We also develop a strongly adaptive variant, which satisfies both guarantees on every time interval. The Defensive Booster is very efficient: it accesses just one weak-class learner, whereas the prior online boosting methods we compare against maintain large weak-learner ensembles. Experiments on synthetic and real data streams demonstrate its strong predictive performance (sometimes substantially improving over all prior baselines) coupled with orders-of-magnitude faster runtime.

Figures

Figures reproduced from arXiv: 2608.13554 by the authors.

Figure 1
Figure 1. Empirical preview (details in Section 6). Top: For each of four naturally ordered real binary streams, each method’s Brier loss is divided by the smallest loss among the methods shown. The Defensive Booster has the lowest Brier loss on Electricity and Occupancy, where it substantially outperforms even the 400-learner aggregator. It essentially ties the aggregator and OGB on Airlines and is within .0010 of the aggreg… view at source ↗
Figure 2
Figure 2. The two baseline families excel on complementary streams, and the Defensive Booster [PITH_FULL_IMAGE:figures/full_fig_p017_2.png] view at source ↗
Figure 3
Figure 3. Multiaccuracy, self-orthogonality, and hard-core diagnostics on the random-label mixture [PITH_FULL_IMAGE:figures/full_fig_p018_3.png] view at source ↗
Figures from the paper (11 more)
Figure 4
Figure 4. Figure 4: Prediction quality as a function of the total number of maintained weak learners on the [PITH_FULL_IMAGE:figures/full_fig_p018_4.png]
Figure 5
Figure 5. Figure 5: Randomized classification error on the streams satisfying the smooth weak-learning [PITH_FULL_IMAGE:figures/full_fig_p036_5.png]
Figure 6
Figure 6. Figure 6: Brier loss on the three linear streams. Top left: on linear span, the span comparator [PITH_FULL_IMAGE:figures/full_fig_p037_6.png]
Figure 7
Figure 7. Figure 7: Online BBM’s specified hard output versus its normalized raw vote, averaged over 20 [PITH_FULL_IMAGE:figures/full_fig_p039_7.png]
Figure 8
Figure 8. Figure 8: One-at-a-time parameter sensitivity on the binary aggregation stream. The dashed [PITH_FULL_IMAGE:figures/full_fig_p040_8.png]
Figure 9
Figure 9. Figure 9: One-at-a-time parameter sensitivity on the complete Electricity (top) and Occupancy [PITH_FULL_IMAGE:figures/full_fig_p040_9.png]
Figure 10
Figure 10. Figure 10: Cumulative average Brier loss on the real-data streams. Top left: on Bank Marketing, [PITH_FULL_IMAGE:figures/full_fig_p042_10.png]
Figure 11
Figure 11. Figure 11: Cumulative mean squared error after the common chronological initialization prefix. [PITH_FULL_IMAGE:figures/full_fig_p045_11.png]
Figure 12
Figure 12. Figure 12: Local performance of the basic and strongly adaptive Defensive Boosters on all four [PITH_FULL_IMAGE:figures/full_fig_p047_12.png]
Figure 13
Figure 13. Figure 13: Local Brier loss on the abrupt and incremental-gradual INSECTS streams. Curves are [PITH_FULL_IMAGE:figures/full_fig_p048_13.png]
Figure 14
Figure 14. Figure 14: Evolution of local hard-core witnesses for the strongly adaptive Defensive Booster on the [PITH_FULL_IMAGE:figures/full_fig_p049_14.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

73 extracted references · 59 canonical work pages

  1. [1]

    Proceedings of the 36th Annual IEEE Symposium on Foundations of Computer Science , pages =

    Impagliazzo, Russell , title =. Proceedings of the 36th Annual IEEE Symposium on Foundations of Computer Science , pages =. 1995 , publisher =. doi:10.1109/SFCS.1995.492584 , url =

    arXiv 1995

  2. [2]

    Advances in Neural Information Processing Systems , volume =

    Bastani, Osbert and Gupta, Varun and Jung, Christopher and Noarov, Georgy and Ramalingam, Ramya and Roth, Aaron , title =. Advances in Neural Information Processing Systems , volume =. 2022 , editor =. doi:10.52202/068431-2129 , url =

    work page doi:10.52202/068431-2129 2022

  3. [3]

    arXiv preprint arXiv:2506.11848 , year =

    Perdomo, Juan Carlos and Recht, Benjamin , title =. arXiv preprint arXiv:2506.11848 , year =. 2506.11848 , archivePrefix =

    arXiv

  4. [4]

    Proceedings of the 42nd International Conference on Machine Learning , pages =

    Noarov, Georgy and Ramalingam, Ramya and Roth, Aaron and Xie, Stephan , title =. Proceedings of the 42nd International Conference on Machine Learning , pages =. 2025 , editor =

    work page 2025

  5. [5]

    A Pseudorandom Generator from Any One-Way Function , journal =

    H. A Pseudorandom Generator from Any One-Way Function , journal =. 1999 , doi =

    work page 1999

  6. [6]

    Annals of Mathematics , volume =

    Green, Ben and Tao, Terence , title =. Annals of Mathematics , volume =. 2008 , doi =

    work page 2008

  7. [7]

    Electronic Colloquium on Computational Complexity , note =

    Reingold, Omer and Trevisan, Luca and Tulsiani, Madhur and Vadhan, Salil , title =. Electronic Colloquium on Computational Complexity , note =. 2008 , url =

    work page 2008

  8. [8]

    , title =

    Trevisan, Luca and Tulsiani, Madhur and Vadhan, Salil P. , title =. Proceedings of the 24th Annual IEEE Conference on Computational Complexity , pages =. 2009 , publisher =

    work page 2009

Show all 73 references
  1. [9]

    2008 , eprint =

    Reingold, Omer and Trevisan, Luca and Tulsiani, Madhur and Vadhan, Salil , title =. 2008 , eprint =

    2008

  2. [10]

    , title =

    Freund, Yoav and Schapire, Robert E. , title =. Journal of Computer and System Sciences , volume =. 1997 , doi =

    1997

  3. [11]

    and Frean, Marcus R

    Mason, Llew and Baxter, Jonathan and Bartlett, Peter L. and Frean, Marcus R. , title =. Advances in Neural Information Processing Systems 12 , pages =. 1999 , url =

    1999

  4. [12]

    , title =

    Mukherjee, Indraneel and Rudin, Cynthia and Schapire, Robert E. , title =. Journal of Machine Learning Research , volume =. 2013 , url =

    2013

  5. [13]

    , title =

    Niculescu-Mizil, Alexandru and Caruana, Richard A. , title =. Proceedings of the Twenty-First Conference on Uncertainty in Artificial Intelligence , pages =. 2005 , eprint =

    2005

  6. [14]

    Proceedings of Innovations in Computer Science , pages =

    Feldman, Vitaly , title =. Proceedings of Innovations in Computer Science , pages =. 2010 , eprint =

    2010

  7. [15]

    Proceedings of the 32nd International Conference on Machine Learning , pages =

    Beygelzimer, Alina and Kale, Satyen and Luo, Haipeng , title =. Proceedings of the 32nd International Conference on Machine Learning , pages =. 2015 , editor =

    2015

  8. [16]

    Advances in Neural Information Processing Systems 28 , pages =

    Beygelzimer, Alina and Hazan, Elad and Kale, Satyen and Luo, Haipeng , title =. Advances in Neural Information Processing Systems 28 , pages =. 2015 , url =

    2015

  9. [17]

    Proceedings of the 20th International Conference on Artificial Intelligence and Statistics , pages =

    Hu, Hanzhang and Sun, Wen and Venkatraman, Arun and Hebert, Martial and Bagnell, Andrew , title =. Proceedings of the 20th International Conference on Artificial Intelligence and Statistics , pages =. 2017 , editor =

    2017

  10. [18]

    Advances in Neural Information Processing Systems 33 , pages =

    Brukhim, Nataly and Chen, Xinyi and Hazan, Elad and Moran, Shay , title =. Advances in Neural Information Processing Systems 33 , pages =. 2020 , url =

    2020

  11. [19]

    Advances in Neural Information Processing Systems , volume =

    Ghai, Udaya and Singh, Karan , title =. Advances in Neural Information Processing Systems , volume =. 2024 , editor =. doi:10.52202/079017-3190 , eprint =

    2024 doi

  12. [20]

    Revisiting Agnostic Boosting , booktitle =

    da Cunha, Arthur and M. Revisiting Agnostic Boosting , booktitle =. 2025 , editor =. 2503.09384 , archivePrefix =

    2025

  13. [21]

    Sample-Near-Optimal Agnostic Boosting with Improved Running Time , booktitle =

    da Cunha, Arthur and H. Sample-Near-Optimal Agnostic Boosting with Improved Running Time , booktitle =. 2026 , editor =. 2601.11265 , archivePrefix =

    2026

  14. [22]

    Multicalibration: Calibration for the (Computationally-Identifiable) Masses , booktitle =

    H. Multicalibration: Calibration for the (Computationally-Identifiable) Masses , booktitle =. 2018 , url =

    2018

  15. [23]

    and Ghorbani, Amirata and Zou, James , title =

    Kim, Michael P. and Ghorbani, Amirata and Zou, James , title =. Proceedings of the 2019 AAAI/ACM Conference on AI, Ethics, and Society , pages =. 2019 , publisher =. doi:10.1145/3306618.3314287 , url =

    2019

  16. [24]

    and Reingold, Omer and Rothblum, Guy N

    Dwork, Cynthia and Kim, Michael P. and Reingold, Omer and Rothblum, Guy N. and Yona, Gal , title =. Proceedings of the 53rd Annual ACM SIGACT Symposium on Theory of Computing , pages =. 2021 , publisher =. doi:10.1145/3406325.3451064 , url =

    2021

  17. [25]

    and Roth, Aaron , title =

    Gupta, Varun and Jung, Christopher and Noarov, Georgy and Pai, Mallesh M. and Roth, Aaron , title =. 13th Innovations in Theoretical Computer Science Conference , volume =. 2022 , doi =

    2022

  18. [26]

    13th Innovations in Theoretical Computer Science Conference , volume =

    Gopalan, Parikshit and Kalai, Adam Tauman and Reingold, Omer and Sharan, Vatsal and Wieder, Udi , title =. 13th Innovations in Theoretical Computer Science Conference , volume =. 2022 , doi =

    2022

  19. [27]

    and Reingold, Omer and Wieder, Udi , title =

    Gopalan, Parikshit and Hu, Lunjia and Kim, Michael P. and Reingold, Omer and Wieder, Udi , title =. 14th Innovations in Theoretical Computer Science Conference , volume =. 2023 , doi =

    2023

  20. [28]

    and Reingold, Omer , title =

    Gopalan, Parikshit and Kim, Michael P. and Reingold, Omer , title =. Advances in Neural Information Processing Systems , volume =

  21. [29]

    Proceedings of the 40th International Conference on Machine Learning , pages =

    Globus-Harris, Ira and Harrison, Declan and Kearns, Michael and Roth, Aaron and Sorrell, Jessica , title =. Proceedings of the 40th International Conference on Machine Learning , pages =. 2023 , url =

    2023

  22. [30]

    Proceedings of the 40th International Conference on Machine Learning , pages =

    Noarov, Georgy and Roth, Aaron , title =. Proceedings of the 40th International Conference on Machine Learning , pages =. 2023 , editor =

    2023

  23. [31]

    Complexity-Theoretic Implications of Multicalibration , booktitle =

    Casacuberta, S. Complexity-Theoretic Implications of Multicalibration , booktitle =. 2024 , publisher =. doi:10.1145/3618260.3649748 , url =

    2024

  24. [32]

    Proceedings of the 2024 Annual ACM-SIAM Symposium on Discrete Algorithms , pages =

    Garg, Sumegha and Jung, Christopher and Reingold, Omer and Roth, Aaron , title =. Proceedings of the 2024 Annual ACM-SIAM Symposium on Discrete Algorithms , pages =. 2024 , doi =

    2024

  25. [33]

    arXiv preprint arXiv:2505.20885 , year =

    Luo, Haipeng and Senapati, Spandan and Sharan, Vatsal , title =. arXiv preprint arXiv:2505.20885 , year =. 2505.20885 , archivePrefix =

    arXiv

  26. [34]

    How Global Calibration Strengthens Multiaccuracy , booktitle =

    Casacuberta, S. How Global Calibration Strengthens Multiaccuracy , booktitle =. 2025 , publisher =. doi:10.1109/FOCS63196.2025.00063 , url =

    2025

  27. [35]

    arXiv preprint arXiv:2507.09683 , year =

    Kearns, Michael and Roth, Aaron and Ryu, Emily , title =. arXiv preprint arXiv:2507.09683 , year =. 2507.09683 , archivePrefix =

  28. [36]

    Proceedings of the 53rd Annual ACM SIGACT Symposium on Theory of Computing , pages =

    Qiao, Mingda and Valiant, Gregory , title =. Proceedings of the 53rd Annual ACM SIGACT Symposium on Theory of Computing , pages =. 2021 , doi =

    2021

  29. [37]

    Proceedings of the 57th Annual ACM Symposium on Theory of Computing , pages =

    Dagan, Yuval and Daskalakis, Constantinos and Fishelson, Maxwell and Golowich, Noah and Kleinberg, Robert and Okoroafor, Princewill , title =. Proceedings of the 57th Annual ACM Symposium on Theory of Computing , pages =. 2025 , doi =

    2025

  30. [38]

    Proceedings of the 42nd International Conference on Machine Learning , pages =

    Ghuge, Rohan and Muthukumar, Vidya and Singla, Sahil , title =. Proceedings of the 42nd International Conference on Machine Learning , pages =. 2025 , url =

    2025

  31. [39]

    Proceedings of Thirty Ninth Conference on Learning Theory , pages =

    Hu, Lunjia and Luo, Haipeng and Senapati, Spandan and Sharan, Vatsal , title =. Proceedings of Thirty Ninth Conference on Learning Theory , pages =. 2026 , editor =

    2026

  32. [40]

    arXiv preprint arXiv:2601.05245 , year =

    Collina, Natalie and Lu, Jiuyao and Noarov, Georgy and Roth, Aaron , title =. arXiv preprint arXiv:2601.05245 , year =. 2601.05245 , archivePrefix =

    arXiv

  33. [41]

    arXiv preprint arXiv:2605.09273 , year =

    Huang, Zhiming and Morgenstern, Jamie and Roth, Aaron and Zhang, Claire Jie , title =. arXiv preprint arXiv:2605.09273 , year =. 2605.09273 , archivePrefix =

    arXiv

  34. [42]

    arXiv preprint arXiv:2604.19592 , year =

    Farina, Gabriele and Perdomo, Juan Carlos , title =. arXiv preprint arXiv:2604.19592 , year =. 2604.19592 , archivePrefix =

    arXiv

  35. [43]

    Proceedings of the Tenth International Workshop on Artificial Intelligence and Statistics , pages =

    Vovk, Vladimir and Takemura, Akimichi and Shafer, Glenn , title =. Proceedings of the Tenth International Workshop on Artificial Intelligence and Statistics , pages =. 2005 , editor =

    2005

  36. [44]

    2007 , month = aug, eprint =

    Vovk, Vladimir , title =. 2007 , month = aug, eprint =

    2007

  37. [45]

    arXiv preprint cs/0506007 , year =

    Vovk, Vladimir and Nouretdinov, Ilia and Takemura, Akimichi and Shafer, Glenn , title =. arXiv preprint cs/0506007 , year =. cs/0506007 , archivePrefix =

    arXiv

  38. [46]

    2003 , url =

    Zinkevich, Martin , title =. 2003 , url =

    2003

  39. [47]

    and Williamson, Robert C

    Kivinen, Jyrki and Smola, Alexander J. and Williamson, Robert C. , title =. IEEE Transactions on Signal Processing , volume =. 2004 , doi =

    2004

  40. [48]

    , title =

    Savage, Leonard J. , title =. Journal of the American Statistical Association , volume =. 1971 , doi =

    1971

  41. [49]

    , title =

    Gneiting, Tilmann and Raftery, Adrian E. , title =. Journal of the American Statistical Association , volume =. 2007 , doi =

    2007

  42. [50]

    Proceedings of the Eighth International Workshop on Artificial Intelligence and Statistics , pages =

    Online Bagging and Boosting , author =. Proceedings of the Eighth International Workshop on Artificial Intelligence and Statistics , pages =. 2001 , editor =

    2001

  43. [51]

    Proceedings of the 29th International Conference on Machine Learning , pages =

    An Online Boosting Algorithm with Theoretical Justifications , author =. Proceedings of the 29th International Conference on Machine Learning , pages =. 2012 , editor =

    2012

  44. [52]

    Proceedings of the 32nd International Conference on Machine Learning , pages =

    Optimal and Adaptive Algorithms for Online Boosting , author =. Proceedings of the 32nd International Conference on Machine Learning , pages =. 2015 , editor =

    2015

  45. [53]

    Proceedings of the 38th International Conference on Machine Learning , pages =

    Boosting for Online Convex Optimization , author =. Proceedings of the 38th International Conference on Machine Learning , pages =. 2021 , editor =

    2021

  46. [54]

    Information and Computation , volume =

    Boosting a Weak Learning Algorithm by Majority , author =. Information and Computation , volume =. 1995 , doi =

    1995

  47. [55]

    Journal of Machine Learning Research , volume =

    Smooth Boosting and Learning with Malicious Noise , author =. Journal of Machine Learning Research , volume =. 2003 , url =

    2003

  48. [56]

    Journal of Machine Learning Research , volume =

    Optimally-Smooth Adaptive Boosting and Application to Agnostic Learning , author =. Journal of Machine Learning Research , volume =. 2003 , url =

    2003

  49. [57]

    Machine Learning , volume =

    Boosting and Hard-Core Set Construction , author =. Machine Learning , volume =. 2003 , doi =

    2003

  50. [58]

    Proceedings of the Twentieth Annual ACM-SIAM Symposium on Discrete Algorithms , pages =

    The Uniform Hardcore Lemma via Approximate Bregman Projections , author =. Proceedings of the Twentieth Annual ACM-SIAM Symposium on Discrete Algorithms , pages =. 2009 , publisher =. doi:10.1137/1.9781611973068.129 , url =

    2009 doi

  51. [59]

    Algorithmic Learning Theory, 16th International Conference, ALT 2005 , series =

    Defensive Forecasting for Linear Protocols , author =. Algorithmic Learning Theory, 16th International Conference, ALT 2005 , series =. 2005 , doi =

    2005

  52. [60]

    Machine Learning , volume =

    Improved Second-Order Bounds for Prediction with Expert Advice , author =. Machine Learning , volume =. 2007 , doi =

    2007

  53. [61]

    Theoretical Computer Science , volume =

    Scale-Free Online Learning , author =. Theoretical Computer Science , volume =. 2018 , doi =

    2018

  54. [62]

    Bank Marketing , year =

    Moro, S. Bank Marketing , year =. doi:10.24432/C5K306 , url =

    doi

  55. [63]

    Massive Online Analysis Datasets , year =

  56. [64]

    1999 , url =

    Harries, Michael , title =. 1999 , url =

    1999

  57. [65]

    and Feldheim, V

    Candanedo, Luis M. and Feldheim, V. Accurate Occupancy Detection of an Office Room from Light, Temperature, Humidity and. Energy and Buildings , volume =. 2016 , doi =

    2016

  58. [66]

    2017 , howpublished =

    Candanedo, Luis , title =. 2017 , howpublished =. doi:10.24432/C5VC8G , url =

    2017 doi

  59. [67]

    2013 , howpublished =

    Fanaee-T, Hadi , title =. 2013 , howpublished =. doi:10.24432/C5W894 , url =

    2013 doi

  60. [68]

    2019 , howpublished =

    Hogue, John , title =. 2019 , howpublished =. doi:10.24432/C5X60B , url =

    2019 doi

  61. [69]

    , title =

    Freund, Yoav and Schapire, Robert E. , title =. Proceedings of the Ninth Annual Conference on Computational Learning Theory , pages =. 1996 , publisher =. doi:10.1145/238061.238163 , url =

    1996

  62. [70]

    Proceedings of the 32nd International Conference on Machine Learning , pages =

    Daniely, Amit and Gonen, Alon and Shalev-Shwartz, Shai , title =. Proceedings of the 32nd International Conference on Machine Learning , pages =. 2015 , editor =

    2015

  63. [71]

    Proceedings of the 27th Conference on Learning Theory , pages =

    Gaillard, Pierre and Stoltz, Gilles and van Erven, Tim , title =. Proceedings of the 27th Conference on Learning Theory , pages =. 2014 , editor =

    2014

  64. [72]

    Proceedings of the 37th International Conference on Machine Learning , pages =

    Cutkosky, Ashok , title =. Proceedings of the 37th International Conference on Machine Learning , pages =. 2020 , editor =

    2020

  65. [73]

    Souza, Vinicius M. A. and dos Reis, Denis M. and Maletzke, Andr. Challenges in Benchmarking Stream Learning Algorithms with Real-world Data , journal =. 2020 , doi =

    2020

Pith tools

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