Pith. sign in

REVIEW 2 major objections 4 minor 1 cited by

Simple and optimal high-probability bounds for strongly-convex stochastic gradient descent

T0 review · 2 major / 4 minor · reviewed 2026-08-14 · deepseek-v4-flash

Pith's one-line read A non-uniform average of SGD iterates reaches the optimal 1/T convergence rate with probability 1-δ, not just in expectation.

desk verdict Solid upper bound for non-uniform averaging SGD; the lower bound misses the theorem's assumptions, so the optimality claim is not yet established. read the letter →

arxiv 1909.00843 v1 pith:S5U3EE5H submitted 2019-09-02 cs.LG math.OCstat.ML

classification cs.LGmath.OCstat.ML
keywords stochasticgradientdescentstronglyconvexnon-smoothoptimizationhigh-probabilityboundsFreedman'sinequalitynon-uniformaveragingmartingaleconcentrationoptimalconvergencerate
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 establishes that a very simple variant of stochastic gradient descent—run ordinary projected SGD, then return the iterates averaged with weights proportional to the iteration number—has error $O(\log(1/\delta)/T)$ with probability at least $1-\delta$ on nonsmooth, strongly convex, Lipschitz problems. That rate matches the optimal $O(1/T)$ convergence already known to hold in expectation, and a matching lower bound shows the $\log(1/\delta)$ factor cannot be removed. The practical upshot is that a single run of SGD can carry a rigorous high-probability guarantee, without running many trials or knowing the stopping time in advance. The proof rests on a martingale whose total conditional variance is bounded by a linear function of the martingale itself, making the tail bound sharp. Experiments on regularized SVM problems indicate the weighted average performs as well as suffix averaging and better than the final iterate or uniform average.

What carries the argument

The proof is carried by a specialized martingale concentration bound: the generalized Freedman inequality, which controls a martingale by its total conditional variance. The relevant martingale has increments $d_t=t\langle \hat z_t, x_t-x^*\rangle$, where $\hat z_t$ is the oracle noise, and its total conditional variance is $V_T=\sum_{t=1}^T t^2\|x_t-x^*\|^2$. The crucial step (Lemma 4.3) shows that, with the step sizes $\eta_t=2/(\mu(t+1))$, this variance can itself be bounded by a linear combination $\sum_i \alpha_i d_i+\beta$ of the martingale increments, with $\alpha_i=O(T/\mu)$ and $\beta=O(L^2T^2/\mu^2)$. This self-bounding, 'chicken-and-egg' structure lets the inequality produce a tail estimate with only a $\log(1/\delta)$ factor instead of the weaker $\sqrt{T}$ factor that generic martingale arguments would give. The weights $t/(T(T+1)/2)$ and step sizes are chosen so that summing the strong-convexity inequalities telescopes, leaving only the noise term to be controlled.

What would settle it

Run Algorithm 1 on $f(x)=\|x\|^2/2$ with noise drawn uniformly from the unit sphere (so $\|\hat z_t\|\le1$ and the theorem's assumptions hold), record the empirical $\delta$-quantile of $f(\sum_t\gamma_tx_t)-f(x^*)$ over many trials; if for any fixed constant $C$ the quantile exceeds $C\log(1/\delta)/T$ as $T$ grows, the upper bound is false. For the lower bound, compute the exact distribution of the weighted average under the paper's Bernoulli oracle on $x^2/2$ and check whether its tail dominates $\log(1/\delta)/(9T)$ with probability $\delta$.

Watch

Extended reading notes

Core claim

The paper's central claim is Theorem 3.1: for a $\mu$-strongly convex, $L$-Lipschitz function and a stochastic gradient oracle whose noise satisfies $\|\hat z_t\|\le 1$ almost surely, projected SGD with step size $\eta_t=2/(\mu(t+1))$ and averaging weights $\gamma_t=t/(T(T+1)/2)$ satisfies $f(\sum_{t=1}^T \gamma_t x_t)-f(x^*) = O((L\log(1/\delta)+L^2/\mu)/T)$ with probability at least $1-\delta$. A companion lower bound, Claim 3.3, constructs a zero-mean Bernoulli oracle on $f(x)=x^2/2$ for which the same averaged output has error at least $\log(1/\delta)/(9T)$ with probability at least $\delta$, so the $\log(1/\delta)$ factor is necessary. Together these show that the simple, online-computable non-uniform average has the optimal high-probability convergence rate, not merely the optimal expected rate.

Load-bearing premise

The entire argument assumes the oracle noise has bounded or at least subgaussian tails; if the noise is heavy-tailed, the martingale variance bound collapses and the $O(\log(1/\delta)/T)$ guarantee may fail.

Editorial extensions

If this is right

  • A single run of SGD with the non-uniform average gives a rigorous $O(\log(1/\delta)/T)$ error bound with probability $1-\delta$, so users can trust one trial instead of averaging many.
  • The $\log(1/\delta)$ factor is unavoidable: no algorithm of this form can beat the $\log(1/\delta)/T$ high-probability rate on nonsmooth strongly convex problems under bounded noise.
  • Non-uniform averaging matches suffix averaging's high-probability guarantee while remaining fully online, with no need to know the time horizon in advance.
  • The subgaussian extension (Theorem C.3) covers Gaussian noise with the same $O(\log(1/\delta)/T)$ rate, replacing $\|\hat z_t\|\le1$ by a $\kappa$-subgaussian assumption.
  • On regularized SVM problems, the experiments show the weighted average and suffix average dominate the final iterate and uniform average in both typical error and spread.

Reading between the lines

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

  • The paper does not prove it, but the same variance self-binding argument likely extends to any output weights within a constant factor of $t$, such as a sliding-window suffix average, which would give practitioners a larger family of simple high-probability averaging schemes.
  • A natural testable extension is the heavy-tailed noise regime: if the oracle noise has only finite variance or power-law tails, one would expect the $\log(1/\delta)/T$ guarantee to degrade, and truncating the stochastic gradient before averaging might restore it.
  • The lower bound's Bernoulli oracle suggests that the $\log(1/\delta)$ factor is intrinsic to subgaussian noise itself; proving a similar lower bound for any first-order randomized method, rather than for this specific averaging scheme, would make that intuition a theorem.
  • The high-probability bound is stated for a fixed horizon $T$; combining it with a union bound over horizons could yield a stopping rule with a guaranteed $\log(1/\delta)\log(T)/T$ rate when $T$ is chosen adaptively.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

2 major / 4 minor

Summary. This paper analyzes stochastic gradient descent with the non-uniform averaging scheme of Lacoste-Julien et al. for nonsmooth strongly convex functions. The main theoretical result is a high-probability upper bound of O((L log(1/δ) + L^2/μ)/T) on the function error, under the assumption that the oracle noise is almost surely bounded by 1. The proof uses a generalized Freedman inequality to bound the martingale formed by the noise terms, together with deterministic bounds adapted from Rakhlin et al. The paper also claims a matching lower bound of Ω(log(1/δ)/T), extends the upper bound to subgaussian noise, and reports experiments comparing averaging schemes.

Significance. If the upper bound is correct, the paper provides a simple algorithm with a high-probability O(log(1/δ)/T) guarantee, which is optimal in T for fixed δ. The proof is clear and the use of the generalized Freedman inequality is appropriate. However, the advertised matching lower bound is not proved for the algorithm and noise model analyzed in the upper bound, so the optimality claim is not currently established. The upper bound and its proof are nevertheless a useful contribution, and the subgaussian extension is a reasonable addition.

major comments (2)
  1. [5, Claim 3.3 and Remark 5.1] The lower bound is proved for step size η_t = 1/(t+1), whereas Algorithm 1 and Theorem 3.1 use η_t = 2/(μ(t+1)) (with μ = 1 for f(x) = x^2/2). Claim 3.3 states it runs Algorithm 1 with these step sizes, which is inconsistent. Remark 5.1 acknowledges the modification and asserts without proof that it is nonessential. As a result, the lower bound does not apply to the algorithm whose upper bound is proved, and the paper's claim of a matching lower bound in the abstract and introduction is unsupported.
  2. [5, lower-bound oracle and Claim 5.2 vs. Theorem 3.1] The oracle used in the lower bound produces noise ẑ_t = ((T+1)/(T-t)) X_t for T/2 < t ≤ 3T/4, which can have magnitude about 4, violating the bounded-noise assumption ‖ẑ_t‖ ≤ 1 used in Theorem 3.1 and in the proof of Lemma 4.1. Claim 5.2 only assumes |ẑ_t| ≤ 6. Thus the lower bound is proved outside the hypotheses of the upper bound, so it does not demonstrate optimality of Algorithm 1 in the setting where the upper bound holds.
minor comments (4)
  1. [A, proof of Lemma 4.5] In Eq. (4), the second summand is written with ‖ĝ_t‖² but the index of summation is i; this appears to be a typo for ‖ĝ_i‖².
  2. [C.2, Lemma C.10] The statement of Lemma C.10 appears to have a typo: the condition should likely be E[exp(X²/κ²)] ≤ 2 (matching the paper's definition of subgaussian) rather than E[exp(X/κ²)] ≤ 2. As written, the lemma does not follow from the cited proposition.
  3. [4, proof of Theorem 3.1] The telescoping step states 'the right-hand side telescopes' after omitting the non-positive term involving ‖x_{T+1}-x*‖²; this is correct but could be stated explicitly to help the reader.
  4. [6, experiments] The abstract claims the non-uniform averaging strategy has 'smaller variance'; the experimental plots in Figure 1 only show trial clouds, so this claim is not quantified. A quantitative variance comparison would strengthen the presentation.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the high-probability upper bound follows from a general martingale inequality plus self-contained lemmas, and the lower-bound mismatch is a validity concern rather than a circular reduction.

full rationale

The paper's central upper bound (Theorem 3.1) is not derived from the result it aims to prove. Its proof uses the Generalized Freedman Inequality (Theorem 2.2), quoted from the authors' prior work Harvey et al. [2018], but that theorem is a general martingale tail bound that does not assume the SGD error bound. The remaining technical ingredients, Lemmas 4.4 and 4.5, are adapted from Rakhlin et al. [2012] and proven in the appendix, and Lemma 4.3 is an explicit decomposition bounding the total conditional variance by a linear combination of martingale increments. No fitted parameter is renamed as a prediction: the upper-bound constants are explicit in the theorem statement, and the proof does not tune a parameter to the target function or to the claimed bound. The lower bound (Claim 3.3) is an adversarial oracle construction on f(x)=x^2/2; it does not fit a parameter to force a specific output, but rather constructs a noise sequence so that the averaged iterate is a Bernoulli average. That construction does use a step size eta_t=1/(t+1) rather than Algorithm 1's eta_t=2/(mu(t+1)), and it allows |z_t|<=6 rather than the |z_t|<=1 used in Theorem 3.1. This is a mismatch between the lower-bound setting and the upper-bound assumptions, and it undermines the advertised optimality claim as a matter of theorem applicability, but it is not a circular derivation: the lower bound does not assume the upper bound, and the upper bound is not defined in terms of the lower bound. The paper's self-citation is to an independent probabilistic tool, not to the paper's own conclusion, so the circularity burden is not met.

Assumptions & free parameters 1 free parameters · 5 assumptions · 0 invented entities

No fitted constants or new physical entities appear in this paper; all constants are universal. The step-size schedule eta_t=2/(mu(t+1)) is adopted from Lacoste-Julien et al. and Rakhlin et al., and the lower-bound oracle is a mathematical construction, not a fitted model. The main ingredients imported from outside are the strong-convexity/Lipschitz problem assumptions, the Generalized Freedman inequality from the same group's prior work, and standard concentration tools.

free parameters (1)
  • step-size schedule eta_t = 2/(mu(t+1)) = 2/(mu(t+1))
    Chosen from prior SGD theory to make the telescoping distance terms sum; it appears in Algorithm 1 and Theorem 3.1, while Claim 3.3 uses eta_t=1/(t+1), so the lower bound does not use the same parameter choice.
assumptions (5)
  • domain assumption Oracle noise z_t has zero mean conditional on the past and is either bounded (||z_t|| <= 1 a.s.) or kappa-subgaussian conditioned on F_{t-1}.
    This is assumption (b) of Theorem 3.1 and the defining condition of Theorem C.3; it is what makes the martingale d_t subgaussian in Claim 4.2 and Claim C.9.
  • domain assumption f is mu-strongly convex and L-Lipschitz on a closed convex set X.
    These are the problem-setting assumptions stated in Section 2; they justify Lemma 4.4 (bounded distance to x*) and the telescoping inequalities in the proof of Theorem 3.1.
  • standard math The Generalized Freedman Inequality of Harvey et al. 2018 (Theorem 2.2 and Theorem 3.3) is valid.
    The main proof invokes this external concentration theorem in Lemma 4.1 and in the subgaussian extension (Theorem C.12); it is from a preprint by overlapping authors and is not reproved here.
  • standard math The reverse Chernoff (Klein-Young) lower tail bound is valid for averages of i.i.d. Bernoulli variables.
    Claim 3.3 uses Lemma B.1 to turn a Bernoulli average into the log(1/delta)/T lower bound.
  • standard math Euclidean projection onto X does not increase distance to any point in X.
    Used in the proof of Theorem 3.1 and in Eq. (2) of the appendix to replace y_{t+1} by the projected point x_{t+1}.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Simple and optimal high-probability bounds for strongly-convex stochastic gradient descent." pith.science (2026). https://pith.science/paper/S5U3EE5H

@misc{pith2026190900843,
  author       = {Pith},
  title        = {Pith review of: Simple and optimal high-probability bounds for strongly-convex stochastic gradient descent},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/S5U3EE5H}},
  note         = {Machine review of arXiv:1909.00843}
}
abstract

We consider stochastic gradient descent algorithms for minimizing a non-smooth, strongly-convex function. Several forms of this algorithm, including suffix averaging, are known to achieve the optimal $O(1/T)$ convergence rate in expectation. We consider a simple, non-uniform averaging strategy of Lacoste-Julien et al. (2011) and prove that it achieves the optimal $O(1/T)$ convergence rate with high probability. Our proof uses a recently developed generalization of Freedman's inequality. Finally, we compare several of these algorithms experimentally and show that this non-uniform averaging strategy outperforms many standard techniques, and with smaller variance.

Figures

Figures reproduced from arXiv: 1909.00843 by the authors.

Figure 1
Figure 1. Additional experimental results can be found in the supplementary material (Section D). [PITH_FULL_IMAGE:figures/full_fig_p007_1.png] view at source ↗
Figure 1
Figure 1. Number of effective passes vs. objective value. The first row plots the results using the [PITH_FULL_IMAGE:figures/full_fig_p009_1.png] view at source ↗
Figure 2
Figure 2. Number of effective passes vs. objective value. Figure 2a plots the results for the [PITH_FULL_IMAGE:figures/full_fig_p019_2.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Almost Sure Convergence Rates and Concentration of Stochastic Approximation and Reinforcement Learning with Markovian Noise

    cs.LG 2024-11 accept novelty 8.0 of 10

    First almost sure convergence rates and maximal concentration bounds with exponential tails for general contractive stochastic approximation with Markovian noise, via a diminishing-interval skeleton discretization.

Reference graph

Works this paper leans on

11 extracted references · 6 canonical work pages · cited by 1 Pith paper

  1. [1]

    Freedman

    David A. Freedman. On tail probabilities for martingales. Annals of Probability, 3 0 (1): 0 100--118, 1975

  2. [2]

    Nicholas J. A. Harvey, Christopher Liaw, Yaniv Plan, and Sikander Randhawa. Tight analyses for non-smooth stochastic gradient descent. CoRR, abs/1812.05217, 2018. URL http://arxiv.org/abs/1812.05217

  3. [3]

    Beyond the regret minimization barrier: optimal algorithms for stochastic strongly-convex optimization

    Elad Hazan and Satyen Kale. Beyond the regret minimization barrier: optimal algorithms for stochastic strongly-convex optimization. The Journal of Machine Learning Research, 15 0 (1): 0 2489--2512, 2014

  4. [4]

    Making the last iterate of SGD information theoretically optimal

    Prateek Jain, Dheeraj Nagaraj, and Praneeth Netrapalli. Making the last iterate of SGD information theoretically optimal. arXiv preprint arXiv:1904.12443, 2019

  5. [5]

    Kakade and Ambuj Tewari

    Sham M. Kakade and Ambuj Tewari. On the generalization ability of online strongly convex programming algorithms. In NIPS, pages 801--808, 2008

  6. [6]

    On the number of iterations for D antzig-- W olfe optimization and packing-covering approximation algorithms

    Philip Klein and Neal E Young. On the number of iterations for D antzig-- W olfe optimization and packing-covering approximation algorithms. SIAM Journal on Computing, 44 0 (4): 0 1154--1172, 2015

  7. [7]

    Schmidt, and Francis R

    Simon Lacoste - Julien, Mark W. Schmidt, and Francis R. Bach. A simpler approach to obtaining an O (1/t) convergence rate for the projected stochastic subgradient method. CoRR, abs/1212.2002, 2012. URL http://arxiv.org/abs/1212.2002

  8. [8]

    Concentration inequalities and model selection

    Pascal Massart. Concentration inequalities and model selection. Springer, 2007

Show all 11 references
  1. [9]

    Nesterov and V

    Yu. Nesterov and V. Shikhman. Quasi-monotone subgradient methods for nonsmooth convex minimization. Journal of Optimization Theory and Applications, 165 0 (3): 0 917--940, Jun 2015

  2. [10]

    Making gradient descent optimal for strongly convex stochastic optimization

    Alexander Rakhlin, Ohad Shamir, and Karthik Sridharan. Making gradient descent optimal for strongly convex stochastic optimization. In Proceedings of ICML, 2012

  3. [11]

    High-dimensional probability: An introduction with applications in data science

    Roman Vershynin. High-dimensional probability: An introduction with applications in data science. Cambridge University Press, 2018

Pith tools

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