Pith. sign in

REVIEW 4 major objections 4 minor 84 references

Safe Start: Configuring Optimization Algorithms for Decision-Making under Extreme Risks

T0 review · 4 major / 4 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read Starting an optimizer in a low-risk region suffices to avoid exponential sampling costs in rare-event decision-making, provided the gradient oracle is variance-reduced.

desk verdict A genuinely new finite-time complexity result for rare-event optimization, with a real but clearly-stated assumption gap that doesn't undermine the core. read the letter →

arxiv 2608.09872 v1 pith:QXXAWC65 submitted 2026-08-10 math.OC stat.ME

Henry Lam , Wasin Meesena This is my paper · ORCID
classification math.OCstat.ME MSC 90C1590C2560F1065K05
keywords rare-eventoptimizationsafestartvariancereductionstochasticgradientdescentsamplingcomplexityconditionalvalue-at-risklargedeviationsrobustclassification
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 studies stochastic optimization where a rare catastrophic event contributes to the objective through a risk penalty, and asks how many simulation samples are needed to reach a near-optimal decision as the event becomes rarer. It argues that common stochastic gradient descent, even with adaptively embedded variance reduction, can require exponentially many samples because the objective landscape is so sensitive that every step size either creeps or overshoots. The paper introduces a 'safe start'—initializing at a decision whose risk is no larger, in exponential scale, than the risk at the optimum—and proves that under this condition, together with efficient variance-reduced gradient estimators, there are SGD schemes whose sampling complexity grows sub-exponentially in the rarity level. It also proves the two ingredients are necessary: without a safe start, exponential complexity can occur for every monotone-diminishing-step SGD, and with a safe start but only an ordinary absolute-variance gradient estimator, exponential complexity can still occur. The practical message is that a conservative, low-risk initialization is not a heuristic nicety but a provable requirement for tractable rare-event optimization.

What carries the argument

The carrying object is the initial sublevel set $S(x_0,\lambda,c)=\{x\in\mathcal{X}: F(x;\lambda)\le cF(x_0;\lambda)\}$. A safe start makes the initial objective value $F(x_0;\lambda)$ sub-exponential in $\lambda$, and Lemma 1 translates this into three uniform controls on $S$: a bounded radius to the optimum, a Lipschitz smoothness constant for $F$, and a bound on the stochastic gradient variance. The SGD analysis then performs a joint containment-and-descent argument: because descent keeps the iterate inside $S$, and inside $S$ the smoothness and noise bounds hold, each step can be shown to both stay contained and decrease the objective; summing these decreases along the trajectory yields the sub-exponential sampling complexity. The necessity results are produced by explicit two- and one-dimensional instances where the absence of a safe start, or of relative-error gradient efficiency, forces exponential iteration counts.

What would settle it

Implement the paper's constructive safe-start configuration for the extreme-quantile or CVaR portfolio example with a variance-reduced gradient oracle, and measure the total number of samples needed to reach $F(\hat{x};\lambda)\le(1+\varepsilon)F(x^*(\lambda);\lambda)$ with probability $1-\kappa$ for increasing rarity levels $\lambda$. If the required sample count grows exponentially in $\lambda$ across a sequence of safe starts satisfying the paper's assumptions, then Theorem 1 is false; a polynomial or sub-exponential growth curve would corroborate it.

Watch

Extended reading notes

Core claim

The central claim is that for a family of optimization problems $\mathcal{P}_\lambda$ with rarity level $\lambda\to\infty$, if the risk at the optimum decays as $e^{-I\lambda}$ and the initial point $x_0(\lambda)$ has risk decaying at least as fast, then there exists an SGD configuration—last-iterate or average-iterate, with constant or polynomially decaying step size—that returns a $(1+\varepsilon)$-optimal solution with probability $1-\kappa$ using a number of samples growing only sub-exponentially in $\lambda$. The proof is constructive: for each rarity level it prescribes a step-size schedule, a mini-batch size, and an iteration horizon, tuned against local smoothness and gradient-noise bounds that the safe start controls. The paper further establishes complementary impossibility results: unsafe starts can force exponential complexity for every SGD scheme with monotonically diminishing step sizes, and even safe starts fail if the gradient estimator only satisfies an absolute variance bound instead of a relative-error efficiency bound.

Load-bearing premise

The entire sufficiency theorem assumes that, for every feasible decision, one already has an unbiased estimator of the rare-event risk gradient whose relative error stays sub-exponential as the event becomes rarer; if no such oracle exists, the proof's positive result does not apply.

Editorial extensions

If this is right

  • In rare-event optimization, initialization is part of the algorithm: starting at a low-risk point converts an exponentially hard problem into a sub-exponential one.
  • Adaptive variance reduction is essential, not optional: without it, even a safe start can fail because gradient noise blows up the required batch size.
  • Constant step sizes work as well as polynomial decay for both last-iterate and average-iterate SGD; the safe start does the heavy lifting.
  • For practitioners, the safe-start condition can be checked from the risk level of the initial decision before full optimization: keep $p(x_0)$ at or below the risk level of the target.
  • The theory supplies explicit step-size, mini-batch, and horizon formulas for each rarity level, so the efficient configuration is constructive rather than existential.

Reading between the lines

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

  • A testable practical recipe follows: for any conservative policy—an all-cash portfolio, a uniform allocation, or a classifier biased toward one class—estimate $p(x_0)$ by simulation and use it as a safe start; the theory predicts a wide band of step sizes will converge once variance-reduced gradients are available.
  • The necessity results suggest that adaptive optimizers like Adam may mask, but not remove, the unsafe-start failure in neural-network training; the paper's robust-classification experiment shows Adam can succeed from an unsafe start, but the safe start gives a larger set of working step sizes, an effect worth testing across more architectures.
  • The proof assumes an oracle that is already the hardest part of the problem: a gradient estimator whose relative error is uniformly sub-exponential over the feasible set. If such an oracle is unavailable, safe start alone may not rescue SGD, and building such oracles for general simulators remains open.
  • There is likely a direct connection to chance-constrained optimization: a safe start is computationally equivalent to choosing a feasible or nearly feasible point for the chance constraint, so existing initialization heuristics for chance-constrained programs could be reinterpreted as complexity guarantees under variance-reduced gradients.
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

4 major / 4 minor

Summary. The paper studies stochastic optimization of objectives that include a rare catastrophic event term, F(x;λ)=f(x)+γ(λ)p(x), as λ→∞. It argues that standard adaptive variance-reduction schemes embedded in SGD can incur exponential sampling complexity due to the ultra-sensitive landscape of the tail term, and proposes a 'safe start' condition (Assumption 3: the initial risk p(x0(λ)) decays at least as fast as the optimal risk) to restore sub-exponential complexity. The main positive result, Theorem 1, states that under Assumptions 1–9 — in particular Assumption 2, a uniform relative-error bound on unbiased gradient estimators for p — a safe start together with a suitably tuned SGD scheme from A_{α,last} or A_{α,avg} yields an efficient configuration in the sense of Definition 2. Two necessity results are proved by explicit counterexamples: Theorem 2 shows that without a safe start, even exact gradients lead to exponential complexity; Theorem 3 shows that even with a safe start, replacing Assumption 2 by a standard absolute-variance bound leads to exponential complexity. The paper also reports numerical experiments in VaR estimation, CVaR portfolio optimization, and robust classification.

Significance. If the main theorem is taken together with its assumptions, the paper makes a genuine conceptual contribution: it identifies a clean geometric condition (safe start) that controls the smoothness and noise constants of the rare-event objective, and it gives the first sampling-complexity bridge between rare-event estimation efficiency (relative error) and optimization efficiency. The negative results are concrete and informative: Proposition 3 gives a deterministic two-dimensional instance where an unsafe start defeats all monotone step-size schedules, and Proposition 4 gives a one-dimensional instance where a noise model satisfying an absolute but not a relative error bound defeats every constant-step-size last-iterate SGD even from a safe start. The proofs in the e-companion are, for the most part, internally consistent and the central derivation is not circular: safe start and gradient-oracle efficiency are explicit assumptions rather than fitted outputs.

major comments (4)
  1. [§2.1, Assumption 2 and Eq. (3)] Assumption 2 is the whole engine of the sufficiency result. It postulates, uniformly over all x in the feasible set, an unbiased estimator G_p satisfying E||G_p − ∇p||^2 ≤ σ_p(λ)^2 ||∇p||^2 with log σ_p(λ) = o(λ). The paper explicitly does not construct or verify such an estimator for any nontrivial problem; Section 2.1 states only that the translation from rare-event probability estimation to gradient estimation 'will need to be conducted case-by-case.' Theorem 3 shows that even a mild-looking relaxation (absolute variance bound) destroys the conclusion. Thus the non-vacuous domain of Theorem 1 is unestablished: the result reduces rare-event optimization to an oracle that may be as hard to obtain as the original problem. The manuscript should either provide a nontrivial family of instances and estimators satisfying Assumption 2, or give sufficient structural conditions under which such uniform relative-error gradient estimators exist, or state this as an open problem in the main text rather than only in passing.
  2. [§2.1, Assumption 2, Eq. (3) and Assumption 7] The relative-error bound (3) is required for all x in the feasible set, including points where ∇p(x) = 0. If p has an interior stationary point with p(x) > 0, the right-hand side of (3) is zero, forcing G_p(x,·;λ) = ∇p(x) almost surely at that point. This is a degeneracy that is generally impossible to guarantee for a stochastic oracle and is not discussed. Natural risk models such as p(x) = E[(Z − x)_+] do not have this problem because their gradient is never zero, but the assumptions as stated cover general p. The authors should either restrict the bound to the region {x : ||∇p(x)|| > 0}, or add a separate condition ensuring ∇p never vanishes on the relevant sublevel sets, or justify why the degenerate case is harmless.
  3. [EC.5, Proposition 4, Case 3 and Eq. (EC.5)] The proof of Theorem 3 is presented as a sketch rather than a complete proof, and this matters because Theorem 3 defines the boundary of the positive result. In the intermediate step-size case, the assertion that 'there exists x0 ∈ X0 such that the next iterate lands in Hλ' is made by continuity without verifying that the image of X0 under the deterministic map covers Hλ; the mini-batch probability argument ('unless the batch is exponentially large, with probability at least one half...') is not formalized with a Chernoff or anti-concentration bound; and the 'balanced noise' case is dismissed without a quantitative calculation. The recovery-from-zero paragraph (EC.5) also implicitly assumes η ≥ e^{−0.1λ} to guarantee Ω(e^λ) iterations, but this link is not stated. These gaps should be closed or at least spelled out with explicit inequalities, especially because the theorem is a main contribution.
  4. [§5, Experimental validation] The experiments illustrate the qualitative phenomena but do not validate Assumption 2 for any of the three settings. For example, in the CVaR portfolio experiment of Section 5.2, the adaptive importance-sampling estimator is defined through the tilting parameter θ_t, but the paper does not show that the resulting G_p satisfies the uniform relative-error bound (3) over the simplex for all λ. Similarly, in Section 5.3, the risk term is estimated by crude Monte Carlo, which certainly does not satisfy Assumption 2; the experiment is presented as a demonstration that safe start helps even without variance reduction, which is fine, but the text should clearly separate 'illustrative experiments' from 'verification of the assumptions.' In the Gaussian model of Section 5.2, closed-form calculations are available and would allow a concrete check of Assumption 2; adding such a check would substantially strengthen the paper's applicability claims.
minor comments (4)
  1. [§3, Theorem 1 statement] Theorem 1 states X = R^n, but the body of the theorem and the proof of Proposition 1 use the general convex set X from Assumption 4. The restriction to R^n should either be removed or its role explained.
  2. [Definition 2 and Proposition 1 proof] The definition of efficiency uses log T^{ε,κ}_λ, but the proof of Proposition 1 allows the sample complexity to be zero when the initial point already satisfies the target accuracy. The logarithm of zero is undefined; this edge case should be handled explicitly, for example by defining log 0 as −∞ or by noting that for sufficiently large λ the non-trivial regime applies.
  3. [§5.3, Table 4] The experiment uses Adam, which is not in the analyzed scheme classes A_{α,last} or A_{α,avg} because Adam uses adaptive, data-dependent step sizes. The text acknowledges this, but it would be helpful to state explicitly that Adam's success falls outside the theoretical guarantees and is included only as an empirical observation.
  4. [Throughout] There are minor typographical and notational inconsistencies: 'limsup' appears sometimes as 'lim sup'; the vector in Eq. (1) uses x but the text sometimes uses bold x inconsistently; and in Section 2.1, the sentence about the growth rate of γ has a missing article. These do not affect the mathematics.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: Theorem 1 derives subexponential SGD sampling complexity from the stated safe-start and relative-error oracle assumptions; necessity theorems are proven by explicit counterexamples, and the author self-citations are not load-bearing.

full rationale

The derivation is self-contained in the relevant sense. Theorem 1 takes Assumption 2 (existence of unbiased gradient estimators with relative-error bound sigma_p sub-exponential in lambda) and Assumption 3 (safe start) as premises, not as fitted outputs; the proof then gives explicit mini-batch and horizon constructions (Lemmas 5 and 10) whose sub-exponential sample complexity is derived from the sub-exponential constants of Lemma 2. The sample-complexity conclusion is not identical to the relative-error assumption: it requires the joint containment-descent analysis in Lemmas 3 through 9 to convert per-step variance control into total iteration and batch bounds, so the sufficiency claim has independent content. The necessity results (Theorems 2 and 3) are supported by explicit counterexample instances (Propositions 3 and 4) that satisfy all other assumptions and violate respectively safe start or Assumption 2; these are not imported from the authors' prior work. Citations to He et al. (2024) and Aolaritei et al. (2025), which share an author, are motivational and related-work references and do not carry the proofs. The paper explicitly acknowledges that constructing such estimators is left case-by-case, stating that "this analysis will need to be conducted case-by-case" and that the focus is "assuming we already have in hand efficient gradient estimators"; this is a limitation on applicability, not a circular derivation. No fitted-parameter prediction, self-definitional identity, or uniqueness-importing move is present.

Assumptions & free parameters 0 free parameters · 9 assumptions · 0 invented entities

The central claim rests on nine explicit assumptions, of which Assumption 2 (uniformly efficient gradient estimators) and Assumption 3 (safe start) are the most restrictive. No free parameters are fitted to data; the lower-bound examples use hand-chosen instances but are not fits.

assumptions (9)
  • domain assumption Assumption 1: lim_{λ→∞} (1/λ) log p*(λ) = -I and limsup (1/λ) log γ(λ) = I.
    Standard large-deviations scaling for rare-event problems; defines the rarity level and the balancing weight. Invoked throughout the asymptotic analysis.
  • domain assumption Assumption 2: Unbiased estimators G_f and G_p exist with E||G_f - ∇f||^2 ≤ σ_f^2 and E||G_p - ∇p||^2 ≤ σ_p(λ)^2 ||∇p||^2, with σ_p sub-exponential.
    This is the strongest premise: it assumes the adaptive variance reduction is uniformly efficient over the whole domain, which is essentially the resolution of the curse of circularity. It is an explicit input, not proved.
  • ad hoc to paper Assumption 3 (Safe start): limsup (1/λ) log p(x0(λ)) ≤ -I.
    New initialization condition introduced by this paper; it requires the initial risk to be no larger, in exponential scale, than the optimal risk.
  • domain assumption Assumption 4: f is twice continuously differentiable, convex, L_f-smooth; X is convex, closed, non-empty.
    Standard regularity in first-order stochastic optimization.
  • domain assumption Assumption 5: The sublevel set {x: f(x) ≤ ε_f} is non-empty with finite diameter d_f.
    Coercivity-type condition ensuring f attains a minimizer and bounded sublevel sets.
  • domain assumption Assumption 6: min f = f(˜x*) = 0.
    WLOG normalization of the average-case loss.
  • domain assumption Assumption 7: ||∇p(x)|| ≤ L_{p,1}(1+||x-˜x*||^{β1}) p(x) and ||∇^2 p(x)|| ≤ L_{p,2}(1+||x-˜x*||^{β2}) p(x).
    Relative gradient/Hessian bounds that hold for exponential-type risk functions such as e^{ψ(x)} and CVaR of normal losses.
  • domain assumption Assumption 8: The combined objective F(·;λ) is convex for every λ.
    Guarantees global optimality and is used for distance-to-optimal bounds; it excludes nonconvex neural network training, so the MAGIC experiment is outside the theory.
  • domain assumption Assumption 9: limsup (1/λ) log ||x0(λ) - ˜x*|| ≤ 0.
    Rules out exponentially far initializations; enables sub-exponential growth of initial objective value.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Safe Start: Configuring Optimization Algorithms for Decision-Making under Extreme Risks." pith.science (2026). https://pith.science/paper/QXXAWC65

@misc{pith2026260809872,
  author       = {Pith},
  title        = {Pith review of: Safe Start: Configuring Optimization Algorithms for Decision-Making under Extreme Risks},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/QXXAWC65}},
  note         = {Machine review of arXiv:2608.09872}
}
read the original abstract

We consider stochastic optimization where the goal is not only to optimize an average-case objective, but also to mitigate the occurrence of rare catastrophic events. This problem is motivated by safety-aware decision-making and AI training. We first argue that, in the presence of a simulation model, natural attempts to integrate variance reduction into optimization, even executed in a reasonable adaptive fashion, encounter fundamental challenges in guaranteeing realistic runtime when using common stochastic gradient descent algorithms. This challenge arises from the extreme sensitivity of tail-based objectives with respect to the decision variables, which renders a dichotomic failure of convergence regardless of what step size we select. We offer remedies based on a new notion of safe start that allows for efficient finite-time error control, and show how the sampling complexity scales favorably under the combination of safe start and variance reduction. We illustrate our methodologies on examples in portfolio optimization and robust classification with neural networks.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

84 extracted references · 47 canonical work pages

  1. [1]

    2008 , publisher=

    Portfolio selection: efficient diversification of investments , author=. 2008 , publisher=

    2008

  2. [2]

    Advances in neural information processing systems , volume=

    Scalable end-to-end autonomous vehicle testing via rare-event simulation , author=. Advances in neural information processing systems , volume=

  3. [3]

    Mathematical programming , volume=

    Multi-stage stochastic optimization applied to energy planning , author=. Mathematical programming , volume=. 1991 , publisher=

    1991

  4. [4]

    IEEE Control Systems Magazine , volume=

    Stochastic model predictive control: An overview and perspectives for future research , author=. IEEE Control Systems Magazine , volume=. 2016 , publisher=

    2016

  5. [5]

    2014 , publisher=

    Introduction to stochastic dynamic programming , author=. 2014 , publisher=

    work page 2014

  6. [6]

    2014 , publisher=

    Understanding machine learning: From theory to algorithms , author=. 2014 , publisher=

    2014

  7. [7]

    Deep Learning , author=

  8. [8]

    SIAM review , volume=

    Optimization methods for large-scale machine learning , author=. SIAM review , volume=. 2018 , publisher=

    2018

Show all 84 references
  1. [9]

    IEEE Transactions on Intelligent Transportation Systems , volume=

    Accelerated evaluation of automated vehicles using piecewise mixture models , author=. IEEE Transactions on Intelligent Transportation Systems , volume=. 2017 , publisher=

    2017

  2. [10]

    Journal of banking & finance , volume=

    Conditional value-at-risk for general loss distributions , author=. Journal of banking & finance , volume=. 2002 , publisher=

    2002

  3. [11]

    2017 , publisher=

    First-order methods in optimization , author=. 2017 , publisher=

    2017

  4. [12]

    arXiv preprint arXiv:1412.6980 , year=

    Adam: A method for stochastic optimization , author=. arXiv preprint arXiv:1412.6980 , year=

    arXiv

  5. [13]

    Annals of operations research , volume=

    A tutorial on the cross-entropy method , author=. Annals of operations research , volume=. 2005 , publisher=

    2005

  6. [14]

    European Journal of Operational Research , volume=

    Optimization of computer simulation models with rare events , author=. European Journal of Operational Research , volume=. 1997 , publisher=

    1997

  7. [15]

    arXiv preprint arXiv:1811.07209 , year=

    A statistical approach to assessing neural network robustness , author=. arXiv preprint arXiv:1811.07209 , year=

    arXiv

  8. [16]

    2004 , publisher=

    Introduction to rare event simulation , author=. 2004 , publisher=

    2004

  9. [17]

    Surveys in Operations Research and Management Science , volume=

    State-dependent importance sampling for rare-event simulation: An overview and recent advances , author=. Surveys in Operations Research and Management Science , volume=. 2012 , publisher=

    2012

  10. [18]

    2004 , publisher=

    Monte Carlo methods in financial engineering , author=. 2004 , publisher=

    2004

  11. [19]

    2016 , publisher=

    Simulation and the Monte Carlo method , author=. 2016 , publisher=

    2016

  12. [20]

    Monte Carlo methods in financial engineering , author=

  13. [21]

    , title =

    He, Shengyi and Jiang, Guangxin and Lam, Henry and Fu, Michael C. , title =. Operations Research , volume =. 0 , doi =

  14. [22]

    arXiv preprint arXiv:2504.03560 , year=

    Stochastic Optimization with Optimal Importance Sampling , author=. arXiv preprint arXiv:2504.03560 , year=

  15. [23]

    SIAM Journal on optimization , volume=

    Robust stochastic approximation approach to stochastic programming , author=. SIAM Journal on optimization , volume=. 2009 , publisher=

    2009

  16. [24]

    2009 , publisher=

    Robust optimization , author=. 2009 , publisher=

    2009

  17. [25]

    arXiv preprint arXiv:1109.5647 , year=

    Making gradient descent optimal for strongly convex stochastic optimization , author=. arXiv preprint arXiv:1109.5647 , year=

    arXiv

  18. [26]

    Advances in neural information processing systems , volume=

    Non-asymptotic analysis of stochastic approximation algorithms for machine learning , author=. Advances in neural information processing systems , volume=

  19. [27]

    2017 , doi =

    Beck, Amir , title =. 2017 , doi =

    2017

  20. [28]

    Mathematical Programming , volume=

    Chance-constrained problems and rare events: an importance sampling approach , author=. Mathematical Programming , volume=. 2016 , publisher=

    2016

  21. [29]

    arXiv preprint arXiv:2304.09221 , year=

    Convergence of stochastic gradient descent under a local Lojasiewicz condition for deep neural networks , author=. arXiv preprint arXiv:2304.09221 , year=

    arXiv

  22. [30]

    INFORMS Journal on Optimization , volume=

    First-order algorithms without Lipschitz gradient: A sequential local optimization approach , author=. INFORMS Journal on Optimization , volume=. 2024 , publisher=

    2024

  23. [31]

    Advances in Neural Information Processing Systems , volume=

    Adaptive first-order methods revisited: Convex minimization without lipschitz requirements , author=. Advances in Neural Information Processing Systems , volume=

  24. [32]

    SIAM Journal on Mathematics of Data Science , volume=

    Gradient Descent in the Absence of Global Lipschitz Continuity of the Gradients , author=. SIAM Journal on Mathematics of Data Science , volume=. 2024 , publisher=

    2024

  25. [33]

    Handbooks in operations research and management science , volume=

    Rare-event simulation techniques: An introduction and recent advances , author=. Handbooks in operations research and management science , volume=. 2006 , publisher=

    2006

  26. [34]

    ACM Trans

    Bai, Yuanlu and Huang, Zhiyuan and Lam, Henry and Zhao, Ding , title =. ACM Trans. Model. Comput. Simul. , month =. 2022 , issue_date =. doi:10.1145/3519385 , abstract =

    2022 doi

  27. [35]

    , booktitle=

    Bai, Yuanlu and He, Shengyi and Lam, Henry and Jiang, Guangxin and Fu, Michael C. , booktitle=. Importance Sampling for Rare-Event Gradient Estimation , year=

  28. [36]

    Rare Event Simulation Using Monte Carlo Methods , pages=

    Rare event simulation for queues , author=. Rare Event Simulation Using Monte Carlo Methods , pages=. 2009 , publisher=

    2009

  29. [37]

    ACM Transactions on Modeling and Computer Simulation (TOMACS) , volume=

    Rare-event simulation for neural network and random forest predictors , author=. ACM Transactions on Modeling and Computer Simulation (TOMACS) , volume=. 2022 , publisher=

    2022

  30. [38]

    and Gómez-Villegas, Miguel and Marin, J

    Gómez, E. and Gómez-Villegas, Miguel and Marin, J. , year =. A multivariate generalization of the power exponential family of distributions , volume =. Communications in Statistics-theory and Methods - COMMUN STATIST-THEOR METHOD , doi =

  31. [39]

    Available at SSRN 4447428 , year=

    Efficient Simulation of Polyhedral Expectations with Applications to Finance , author=. Available at SSRN 4447428 , year=

  32. [40]

    Large Deviation Theory , author=

  33. [41]

    2003 , publisher=

    Convex analysis and optimization , author=. 2003 , publisher=

    2003

  34. [42]

    Year , note =

    Stefan Adams , title =. Year , note =

  35. [43]

    INFORMS Journal on Optimization , year=

    First-order algorithms without lipschitz gradient: A sequential local optimization approach , author=. INFORMS Journal on Optimization , year=

  36. [44]

    SIAM Journal on Optimization , volume =

    Beck, Amir and Tetruashvili, Luba , title =. SIAM Journal on Optimization , volume =. 2013 , doi =. https://doi.org/10.1137/120887679 , abstract =

    2013 doi

  37. [45]

    Proceedings of The 24th International Conference on Artificial Intelligence and Statistics , pages =

    Deep Probabilistic Accelerated Evaluation: A Robust Certifiable Rare-Event Simulation Methodology for Black-Box Safety-Critical Systems , author =. Proceedings of The 24th International Conference on Artificial Intelligence and Statistics , pages =. 2021 , editor =

    2021

  38. [46]

    and Nobukawa, Kazutoshi and Pan, Christopher S

    Zhao, Ding and Lam, Henry and Peng, Huei and Bao, Shan and LeBlanc, David J. and Nobukawa, Kazutoshi and Pan, Christopher S. , journal=. Accelerated Evaluation of Automated Vehicles Safety in Lane-Change Scenarios Based on Importance Sampling Techniques , year=

  39. [47]

    Adaptive First-Order Methods Revisited: Convex Minimization without Lipschitz Requirements , volume =

    Antonakopoulos, Kimon and Mertikopoulos, Panayotis , booktitle =. Adaptive First-Order Methods Revisited: Convex Minimization without Lipschitz Requirements , volume =

  40. [48]

    SIAM Journal on Optimization , volume =

    Tong, Shanyin and Subramanyam, Anirudh and Rao, Vishwas , title =. SIAM Journal on Optimization , volume =. 2022 , doi =

    2022

  41. [49]

    Management Science , volume=

    Variance reduction techniques for estimating value-at-risk , author=. Management Science , volume=. 2000 , publisher=

    2000

  42. [50]

    arXiv preprint arXiv:2407.11825 , year=

    Optimization under rare events: scaling laws for linear chance-constrained programs , author=. arXiv preprint arXiv:2407.11825 , year=

  43. [51]

    National Bureau of Standards applied mathematics series , volume=

    Estimation of particle transmission by random sampling , author=. National Bureau of Standards applied mathematics series , volume=. 1951 , publisher=

    1951

  44. [52]

    Applied Mathematics & Optimization , volume=

    Simulation and estimation of extreme quantiles and extreme probabilities , author=. Applied Mathematics & Optimization , volume=. 2011 , publisher=

    2011

  45. [53]

    Mathematical programming , volume=

    Uncertain convex programs: randomized solutions and confidence levels , author=. Mathematical programming , volume=. 2005 , publisher=

    2005

  46. [54]

    2004 , publisher=

    The cross-entropy method: a unified approach to combinatorial optimization, Monte-Carlo simulation and machine learning , author=. 2004 , publisher=

    2004

  47. [55]

    2007 , publisher=

    Stochastic simulation: algorithms and analysis , author=. 2007 , publisher=

    2007

  48. [56]

    Operations Research , volume=

    Multilevel splitting for estimating rare event probabilities , author=. Operations Research , volume=. 1999 , publisher=

    1999

  49. [57]

    Stochastic processes and their applications , volume=

    Splitting for rare event simulation: A large deviation approach to design and analysis , author=. Stochastic processes and their applications , volume=. 2009 , publisher=

    2009

  50. [58]

    Statistics and Computing , volume=

    Markov chain importance sampling with applications to rare event probability estimation , author=. Statistics and Computing , volume=. 2013 , publisher=

    2013

  51. [59]

    Network performance engineering: a handbook on convergent multi-service networks and next generation internet , pages=

    The rare event simulation method RESTART: efficiency analysis and guidelines for its application , author=. Network performance engineering: a handbook on convergent multi-service networks and next generation internet , pages=. 2011 , publisher=

    2011

  52. [60]

    2013 , publisher=

    Modelling extremal events: for insurance and finance , author=. 2013 , publisher=

    2013

  53. [61]

    2015 , publisher=

    Quantitative risk management: concepts, techniques and tools-revised edition , author=. 2015 , publisher=

    2015

  54. [62]

    Operations Research , volume=

    Achieving efficiency in black-box simulation of distribution tails with self-structuring importance samplers , author=. Operations Research , volume=. 2025 , publisher=

    2025

  55. [63]

    Operations Research , volume=

    Adaptive importance sampling for efficient stochastic root finding and quantile estimation , author=. Operations Research , volume=. 2024 , publisher=

    2024

  56. [64]

    2009 , publisher=

    Large deviations techniques and applications , author=. 2009 , publisher=

    2009

  57. [65]

    ACM Transactions on Modeling and Computer Simulation (TOMACS) , volume=

    Asymptotic robustness of estimators in rare-event simulation , author=. ACM Transactions on Modeling and Computer Simulation (TOMACS) , volume=. 2010 , publisher=

    2010

  58. [66]

    Handbook of Quantile Regression , pages=

    Extremal quantile regression , author=. Handbook of Quantile Regression , pages=. 2017 , publisher=

    2017

  59. [67]

    The Annals of Applied Statistics , volume=

    Neural networks for extreme quantile regression with an application to forecasting of flood risk , author=. The Annals of Applied Statistics , volume=. 2024 , publisher=

    2024

  60. [68]

    Asuncion, Arthur and Newman, David and others , year=

  61. [69]

    arXiv preprint arXiv:1412.6572 , year=

    Explaining and harnessing adversarial examples , author=. arXiv preprint arXiv:1412.6572 , year=

    arXiv

  62. [70]

    2017 IEEE symposium on security and privacy (sp) , pages=

    Towards evaluating the robustness of neural networks , author=. 2017 IEEE symposium on security and privacy (sp) , pages=. 2017 , organization=

    2017

  63. [71]

    International Conference on Machine Learning , pages=

    PROVEN: Verifying robustness of neural networks with a probabilistic approach , author=. International Conference on Machine Learning , pages=. 2019 , organization=

    2019

  64. [72]

    Representations and Weak Convergence Methods

    Analysis and approximation of rare events , author=. Representations and Weak Convergence Methods. Series Prob. Theory and Stoch. Modelling , volume=. 2019 , publisher=

    2019

  65. [73]

    Automatica , volume=

    Infinitesimal and finite perturbation analysis for queueing networks , author=. Automatica , volume=. 1983 , publisher=

    1983

  66. [74]

    Management Science , volume=

    Convergence properties of infinitesimal perturbation analysis estimates , author=. Management Science , volume=. 1988 , publisher=

    1988

  67. [75]

    Mathematics and Computers in Simulation , volume=

    The score function approach for sensitivity analysis of computer simulation models , author=. Mathematics and Computers in Simulation , volume=. 1986 , publisher=

    1986

  68. [76]

    Operations Research , volume=

    Sensitivity analysis for simulations via likelihood ratios , author=. Operations Research , volume=. 1989 , publisher=

    1989

  69. [77]

    Communications of the ACM , volume=

    Likelihood ratio gradient estimation for stochastic systems , author=. Communications of the ACM , volume=. 1990 , publisher=

    1990

  70. [78]

    Journal of Optimization Theory and Applications , volume=

    Measure-valued differentiation for Markov chains , author=. Journal of Optimization Theory and Applications , volume=. 2008 , publisher=

    2008

  71. [79]

    ACM Transactions on Modeling and Computer Simulation (TOMACS) , volume=

    Gradient estimation for discrete-event systems by measure-valued differentiation , author=. ACM Transactions on Modeling and Computer Simulation (TOMACS) , volume=. 2010 , publisher=

    2010

  72. [80]

    Operations Research , volume=

    A new unbiased stochastic derivative estimator for discontinuous sample performances with structural parameters , author=. Operations Research , volume=. 2018 , publisher=

    2018

  73. [81]

    2024 Winter Simulation Conference (WSC) , pages=

    Importance sampling for minimization of tail risks: A tutorial , author=. 2024 Winter Simulation Conference (WSC) , pages=. 2024 , organization=

    2024

  74. [82]

    Stochastic Systems , volume=

    Efficient scenario generation for heavy-tailed chance constrained optimization , author=. Stochastic Systems , volume=. 2024 , publisher=

    2024

  75. [83]

    Mathematical methods in stochastic simulation and experimental design: Proceedings of the 2nd st

    Importance sampling for Monte Carlo estimation of quantiles , author=. Mathematical methods in stochastic simulation and experimental design: Proceedings of the 2nd st. petersburg workshop on simulation , pages=. 1996 , organization=

    1996

  76. [84]

    2022 , publisher=

    The fundamentals of heavy tails: Properties, emergence, and estimation , author=. 2022 , publisher=

    2022

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.