Pith. sign in

REVIEW 4 major objections 4 minor 41 references

Stochastic Polynomial Optimization

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

Pith's one-line read A norm perturbation makes sample-average polynomial optimization reliably solvable.

desk verdict A correct tightness theorem for a regularized SAA scheme, but the paper never shows the perturbation tracks the true stochastic optimum, and the numerics hand-pick epsilon. read the letter →

arxiv 1908.05689 v1 pith:FKH6BNLR submitted 2019-08-15 math.OC

classification math.OC MSC 90C1590C2290C3165K05
keywords stochasticpolynomialoptimizationsampleaverageapproximationmomentrelaxationssum-of-squaressemidefiniteprogrammingnonconvexperturbationregularization
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 proposes a perturbation sample average approximation (PSAA) model: instead of minimizing the sample average $f_N(x)$ directly, minimize $f_N(x) + \epsilon\|[x]_{2d}\|$ over the same feasible set, with a small $\epsilon > 0$, and solve it with Moment-SOS relaxations. The paper's central claim is that this perturbation cures a known failure mode of sample average approximation—the sampled problem can be unbounded below even when the true expectation problem is well posed—and that the moment relaxation is tight exactly when the moment matrix of its minimizer has rank one. In that case, the projection of the relaxed minimizer is a global minimizer of the perturbed problem. Numerical experiments on polynomial objectives with uniform, normal, Poisson, Bernoulli, and geometric noise show the perturbed model returning reliable minimizers where the unperturbed sample average model fails. A sympathetic reader would care because this gives nonconvex stochastic polynomial optimization a route through semidefinite programming without assuming convexity or gradient information.

What carries the argument

The central object is the perturbed sample average approximation (PSAA): minimize $f_N(x) + \epsilon\|[x]_{2d}\|$ subject to $g_i(x) \ge 0$, where $[x]_{2d}$ is the vector of all monomials up to degree $2d$ and $\epsilon > 0$ is chosen small. The workhorse is the moment relaxation (3.3), a semidefinite program over truncated moment sequences $y$ with $y_0 = 1$, moment matrix $M_d[y] \succeq 0$, and localizing matrices $L_{g_i}^{(d)}[y] \succeq 0$. The norm term $\epsilon\|y\|$ makes the relaxed objective strictly convex, which forces uniqueness of the minimizer and enables the rank-one characterization. Its dual (3.4) has a clear reading: it finds the largest $\gamma$ such that $f_N - \gamma$ lies in the truncated quadratic module $Q(g)_{2d}$ up to a small polynomial perturbation $p$ with $\|\operatorname{vec}(p)\| \le \epsilon$.

What would settle it

Take a stochastic polynomial optimization with known expectation $f$, generate many independent sample sets, and for each set choose $\epsilon$ by solving (3.5); if the returned points $u$ stay bounded away from the true minimizer $v^*$ as the sample size grows, the practical claim that PSAA recovers reliable minimizers would fail.

Watch

Extended reading notes

Core claim

The paper establishes Theorem 3.2: for $\epsilon > 0$, the moment relaxation (3.3) of the perturbed sample average problem (3.2) is tight if and only if the moment matrix $M_d[y^*]$ of its minimizer has rank one; when this rank-one condition holds, $u = \pi(y^*)$ is a minimizer of (3.2). The proof runs through strict convexity: for $\epsilon > 0$ the objective of (3.3) is strictly convex, so the minimizer is unique, and if the point returned by the relaxation is genuinely a feasible point of (3.2), its moment matrix factors as $[u]_d [u]_d^T$, giving rank one. Conversely, tightness forces the relaxed minimizer to coincide with the lifted point generated by a minimizer of (3.2), whose moment matrix is rank one. The paper also shows (Theorem 3.3) that when the feasible set has nonempty interior, a sufficiently large $\epsilon$ makes both the primal relaxation and its dual feasible with equal optimal values, and it defines the minimal such $\epsilon$ as the solution of an auxiliary semidefinite program (3.5).

Load-bearing premise

The load-bearing premise is that a small parameter $\epsilon$ exists for which the perturbed sample-average problem is a faithful stand-in for the original expectation problem; the paper assumes this closeness without proving it, and the bias introduced by $\epsilon\|[x]_{2d}\|$ is not quantified.

Editorial extensions

If this is right

  • For any instance where the relaxed minimizer has rank-one moment matrix, the method certifies a global minimizer of the perturbed problem (3.2) and extracts it from the first moments.
  • The perturbation regularizes sample-average problems that are unbounded from below: for large enough $\epsilon$, the relaxation always has a minimizer, so degenerate samples no longer make the model unsolvable.
  • The minimal useful $\epsilon$ can be chosen by solving the auxiliary problem (3.5), which gives the smallest norm perturbation making the dual feasible.
  • Because the pipeline is polynomial arithmetic plus semidefinite programming, it applies to nonconvex objectives and to discrete sample distributions without gradient evaluations.

Reading between the lines

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

  • A testable extension is to treat $\epsilon$ as a tuning parameter and measure the bias-variance trade-off across many independent sample realizations; the paper demonstrates individual cases but does not quantify the distribution of the distance $\|u - v^*\|$.
  • The $\epsilon\|y\|$ term acts like a Tikhonov regularizer in moment space, not in decision space; this suggests the method will be most valuable exactly when the unperturbed sample average is ill-posed, and less so when the sample average is already well behaved.
  • One could ask whether the rank-one condition holds asymptotically as $N \to \infty$ and $\epsilon \to 0$ at an appropriate joint rate; the paper does not address this limit, and a positive answer would sharpen the practical promise of the method.
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 proposes a perturbed sample average approximation (PSAA) model for stochastic polynomial optimization, in which the sampled objective f_N(x) is replaced by f_N(x) + ε‖[x]_{2d}‖ over a semialgebraic feasible set K. The authors study properties of this perturbed problem and of its Lasserre-type moment relaxation: Theorem 3.2 characterizes tightness of the relaxation by the rank-one condition on the moment matrix of its minimizer, and Theorem 3.3 asserts solvability for large ε. They present Algorithm 3.1, which chooses a small ε, solves the moment relaxation (3.3), and outputs the projection u of the minimizer. The numerical section compares PSAA with the classical SAA (ε=0) on seven examples with various distributions, reporting that PSAA often returns minimizers close to the true optimizer while the plain SAA may be unsolvable or return poor solutions.

Significance. If the proposed method delivered guaranteed or even well-controlled approximations to the original stochastic problem (1.3), it would be a valuable addition to the toolbox for nonconvex stochastic polynomial optimization. The central theoretical observation, Theorem 3.2, is correct and is proved by a clean argument: for fixed ε, tightness of the moment relaxation is equivalent to rank(M_d[y*]) = 1, in which case the projection is a global minimizer of (3.2). This is a useful structural result. However, the practical significance of the paper hangs on an unsupported bridge between the perturbed sample problem (3.2) and the true problem (1.3), and the paper states no error bound, no convergence rate, and no principled rule for selecting ε. The numerical experiments, though suggestive, use hand-picked values of ε and deterministic or single-path sample constructions. The contribution is therefore best viewed as a heuristic proposal with one clean correctness statement, while the claimed practical advantage over classical SAA is not yet established.

major comments (4)
  1. [Section 3, Eq. (3.2)] The paper never provides a quantitative relationship between the solution set of the perturbed sample problem (3.2) and that of the original stochastic problem (1.3). The uniform-convergence argument quoted in Section 1 applies to the plain sample average f_N converging to f, not to f_N + ε‖[x]_{2d}‖ converging to f. For any fixed ε > 0, the objective in (3.2) approximates f + ε‖[x]_{2d}‖, whose minimizer can differ from the minimizer of f; conversely, too small an ε may leave (3.3) unbounded below. Algorithm 3.1 offers no rule for balancing these two effects. This is a load-bearing gap, because the paper's conclusion that Algorithm 3.1 gives reliable minimizers of (1.3) rests entirely on the unstated assumption that a usable ε interval exists and can be found without prior knowledge of the true solution.
  2. [Theorem 3.3] The proof of Theorem 3.3 asserts that Q(g)_{2d} is a closed cone whenever K has nonempty interior, citing [23, Theorem 3.49]. Nonempty interior of K is by itself not sufficient for closedness of the truncated quadratic module; known closedness results require additional conditions, typically archimedeanity of the full quadratic module. Since this theorem is used to guarantee that (3.3) and (3.4) have optimizers for large ε, the statement needs either a corrected proof under explicit sufficient assumptions or a revised argument that does not depend on this closedness claim.
  3. [Section 4, Examples 4.5-4.6 (Tables 6 and 8)] The numerical results do not support the implicit message that the value ε* defined in (3.5) is a useful 'minimal feasible' perturbation. In Table 6, case III, the relaxation at ε* has |⟨f_N, y*⟩ − f_N(u)| = 5.30e+01, and in Table 8, case II, the corresponding gap is 1.13e+02, showing that the moment relaxation is far from tight at ε*. The successful entries in these tables use larger values (0.1 and 0.05) that are chosen on an ad hoc, case-by-case basis and are not derived from (3.5) or any other stated rule. Thus the experiments demonstrate only that the method can work when ε is tuned after the fact, not that Algorithm 3.1 as written provides a reliable way to select ε.
  4. [Section 4, Examples 4.1-4.7] The experimental design does not substantiate the claimed practical advantage in a statistical sense. In Examples 4.1 and 4.2, sample averages are constructed deterministically to have a prescribed deviation from the expectation, rather than being drawn randomly; in Examples 4.5-4.7, the tables report a single sample path per case. No repeated trials, confidence intervals, or sensitivity analysis over ε are presented. Consequently, the tables establish that the method succeeds on a few hand-picked instances and parameter choices, but they do not constitute evidence that the algorithm reliably outperforms the classical SAA for typical or worst-case sample realizations.
minor comments (4)
  1. [Section 2, paragraph before Eq. (2.4)] The word 'semidefintie' appears to be a typo for 'semidefinite'.
  2. [Section 3, after Eq. (3.3)] The text says 'The set of all y satisfying linear matrix inequalities in (3.3) is just the cone S(g)_{2d}', but the constraint y0 = 1 is also present, so the relevant set is S(g)_{2d} ∩ {y0 = 1}.
  3. [Theorem 3.3, proof] The expression 'for the choice ǫ > ‖f_N − [x]_d^T [x]_d‖' is ambiguous: the left norm is a scalar assigned to ε, while the right-hand side must be the norm of a coefficient vector (or a polynomial norm), not the Euclidean norm of a monomial vector. Please clarify the norm being used.
  4. [Algorithm 3.1, Step 2] The instruction 'If (3.3) is infeasible' is imprecise, since (3.3) is feasible whenever K is nonempty. The intended condition is likely that (3.3) is unbounded below or does not have a minimizer.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the main equivalence is a self-contained fixed-epsilon theorem, and heuristic epsilon selection is a validation gap rather than a circular reduction.

full rationale

The paper's central theoretical result, Theorem 3.2, is a direct mathematical equivalence for any fixed positive epsilon: the moment relaxation (3.3) is tight for the perturbed sample-average problem (3.2) exactly when the minimizer's moment matrix has rank one. The proof unfolds the definition of tightness and uses strict convexity of the perturbed objective to identify the lifted feasible point y* = [pi(y*)]_{2d}, so no fitted parameter or data-dependent quantity is introduced into the derivation. The perturbation parameter epsilon is an input to the model, not an output fitted from the data, and the paper explicitly treats its selection as heuristic. The lack of a quantitative bridge between the perturbed sample-average problem (3.2) and the original expectation problem (1.3), together with the case-by-case choice of epsilon in the numerical experiments, is a genuine correctness and validation concern, but it is not a circular reduction: the theorems do not assume the numerical conclusions they are used to support. The self-citations are to standard flat-extension and convergence results used as external mathematical facts, and they are not load-bearing in a way that forces the target result. Therefore, no specific circular step can be exhibited from the paper's own equations, and the derivation is self-contained with respect to its stated claims.

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

The paper depends on standard polynomial optimization machinery and the distributional assumption on samples. The main ad hoc element is the choice of the perturbation term and its parameter epsilon; the paper does not provide a principled way to set epsilon or to bound the bias it introduces.

free parameters (1)
  • epsilon = chosen heuristically (e.g., 1e-2) or as epsilon* from (3.5)
    The perturbation parameter is introduced ad hoc. Its value controls the tradeoff between solvability and approximation quality, but the paper gives no rule for choosing it optimally, only heuristics.
assumptions (5)
  • standard math Lasserre Moment-SOS hierarchy and its convergence properties (from [18,20,23]).
    Used to solve the polynomial optimization and to derive the dual (3.4).
  • standard math Putinar's Positivstellensatz ([34]).
    Used to assert that positive polynomials on K belong to the quadratic module, though not directly invoked in the proofs.
  • domain assumption Samples xi^(k) are i.i.d. from the distribution of xi; law of large numbers applies.
    Justifies the SAA replacement of the expectation by f_N.
  • domain assumption The feasible set K has nonempty interior (for Theorem 3.3).
    Used to construct an interior point of the cone S(g)_{2d} via a Gaussian measure.
  • ad hoc to paper The perturbation term epsilon ||[x]_{2d}|| is a valid regularizer that preserves the polynomial structure.
    No derivation; it is justified only by numerical success.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Stochastic Polynomial Optimization." pith.science (2026). https://pith.science/paper/FKH6BNLR

@misc{pith2026190805689,
  author       = {Pith},
  title        = {Pith review of: Stochastic Polynomial Optimization},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FKH6BNLR}},
  note         = {Machine review of arXiv:1908.05689}
}
read the original abstract

This paper studies stochastic optimization problems with polynomials. We propose an optimization model with sample averages and perturbations. The Lasserre type Moment-SOS relaxations are used to solve the sample average optimization. Properties of the optimization and its relaxations are studied. Numerical experiments are presented.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

41 extracted references · 40 canonical work pages

  1. [1]

    Bastin, C

    F. Bastin, C. Cirillo and P. Toint Convergence theory for nonconvex stochastic programming with an application to mixed logit. Math. Program., 108(23), pp. 207–234, 2006

  2. [2]

    Ben-Tal and A

    A. Ben-Tal and A. Nemirovski, Lectures on Modern Convex Optimization: Analysis, Algo- rithms, and Engineering Applications , MPS-SIAM Series on Optimization, SIAM, Philadel- phia, 2001

  3. [3]

    Blekherman, P

    G. Blekherman, P. Parrilo and R. Thomas (eds.), Semidefinite optimization and convex al- gebraic geometry, MOS-SIAM Series on Optimization, SIAM, 2013

  4. [4]

    Boyd and L

    S. Boyd and L. Vandenberghe, Convex Optimization , Cambridge University Press, 2004

  5. [5]

    M. Branda. Sample approximation technique for mixed-in teger stochastic programming prob- lems with several chance constraints. Operations Research Letters, 40(3), pp. 207–211, 2012

  6. [6]

    Chung, A course in probability theory , Academic press, 2001

    K. Chung, A course in probability theory , Academic press, 2001

  7. [7]

    Curto and L

    R. Curto and L. Fialkow. Truncated K-moment problems in s everal variables. Journal of Operator Theory, 54(2005), pp. 189-226

  8. [8]

    Fu (eds.), Handbook of simulation optimization , volume 216, Springer, New York, 2015

    M. Fu (eds.), Handbook of simulation optimization , volume 216, Springer, New York, 2015

Show all 41 references
  1. [9]

    Ghadimi and G

    S. Ghadimi and G. Lan. Stochastic first-and zeroth-order methods for nonconvex stochastic programming, SIAM J. Optim. , 23(4), pp. 2341–2368, 2013

  2. [10]

    Ghadimi, G

    S. Ghadimi, G. Lan and H. Zhang. Mini-batch stochastic a pproximation methods for non- convex stochastic composite optimization, Math. Program., 155(1-2), pp. 267–305, 2016

  3. [11]

    Helton and J

    J. Helton and J. Nie. Semidefinite representation of con vex sets. Math. Program., Series A , Vol. 122, No.1, pp. 21-64, 2010

  4. [12]

    Helton and J

    J. Helton and J. Nie, A semidefinite approach for truncat ed K-moment problem, Found. Comput. Math. , 12(6), pp. 851–881, 2012

  5. [13]

    Henrion and J

    D. Henrion and J. Lasserre. Detecting global optimalit y and extracting solutions in Glop- tiPoly. Positive polynomials in control , 293–310, Lecture Notes in Control and Inform. Sci., 312, Springer, Berlin, 2005

  6. [14]

    Henrion, J

    D. Henrion, J. Lasserre, and J. L¨ ofberg. Gloptipoly 3: moments, optimization and semidef- inite programming. Optimization Methods & Software , 24(4-5), pp. 761–779, 2009

  7. [15]

    K. Judd. The law of large numbers with a continuum of iid r andom variables. Journal of Economic theory, 35(1), pp. 19–25, 1985

  8. [16]

    Klerk and M

    E. Klerk and M. Laurent. On the Lasserre hierarchy of sem idefinite programming relaxations of convex polynomial optimization problems, SIAM J. Optim. , 21, pp. 824-832, 2011

  9. [17]

    Kleywegt, A

    A. Kleywegt, A. Shapiro, and T. Mello. The sample averag e approximation method for sto- chastic discrete optimization. SIAM J. Optim. , 12(2), pp. 479–502, 2002

  10. [18]

    Lasserre, Global optimization with polynomials and the problem of moments, SIAM J

    J. Lasserre, Global optimization with polynomials and the problem of moments, SIAM J. Optim., 11, pp. 796–817, 2001

  11. [19]

    Lasserre

    J. Lasserre. Convexity in semi-algebraic geometry and polynomial optimization. SIAM J. Optim., 19 (2009), pp. 1995-2014

  12. [20]

    Lasserre

    J. Lasserre. Introduction to polynomial and semi-algebraic optimizati on, Cambridge Univer- sity Press, Cambridge, 2015

  13. [21]

    Lasserre, M

    J. Lasserre, M. Laurent, and P. Rostalski. Semidefinite characterization and computation of zero-dimensional real radical ideals. Found. Comput. Math. 8, pp. 607–647, 2008

  14. [22]

    M. Laurent. Semidefinite representations for finite var ieties. Math. Program., Vol. 109, pp. 1–26, 2007. STOCHASTIC POLYNOMIAL OPTIMIZATION 17

  15. [23]

    M. Laurent, Sums of squares, moment matrices and optimi zation over polynomials, Emerging Applications of Algebraic Geometry of IMA Volumes in Mathem atics and its Applications , 149, pp. 157–270, Springer, 2009

  16. [24]

    Linderoth, A

    J. Linderoth, A. Shapiro, and S. W right. The empirical b ehavior of sampling methods for stochastic programming. Annals of Operations Research , 142(1):215–241, 2006

  17. [25]

    Luedtke and S

    J. Luedtke and S. Ahmed. A sample approximation approac h for optimization with proba- bilistic constraints. SIAM J. Optim. , 19(2), pp. 674–699, 2008

  18. [26]

    J. Nie. Certifying convergence of Lasserre’s hierarch y via flat truncation. Math. Program., Ser. A, 142 (2013), no. 1-2, pp. 485–510

  19. [27]

    Nie, Polynomial optimization with real varieties, SIAM J

    J. Nie, Polynomial optimization with real varieties, SIAM J. Optim. , 23(3), pp. 1634–1646, 2013

  20. [28]

    J. Nie. Optimality conditions and finite convergence of Lasserre’s hierarchy.Math. Program., Ser. A, 146 (2014), no. 1-2, pp. 97–121

  21. [29]

    Nie, The A-truncated K-moment problem, Found

    J. Nie, The A-truncated K-moment problem, Found. Comput. Math. , 14(6), pp. 1243–1276, 2014

  22. [30]

    J. Nie. Linear optimization with cones of moments and no nnegative polynomials. Math. Pro- gram., Ser. B, 153 (2015), no. 1, pp. 247–274

  23. [31]

    P. Parrilo. Semidefinite Programming relaxations for s emialgebraic problems. Math. Pro- gram., Ser. B 96 (2003), No. 2, pp. 293–320

  24. [32]

    Parrilo and B

    P. Parrilo and B. Sturmfels. Minimizing polynomial fun ctions. Algorithmic and quantitative real algebraic geometry (Piscataway, NJ, 2001) , pp. 83–99, DIMACS Ser. Discrete Math. Theoret. Comput. Sci., 60, Amer. Math. Soc., Providence, RI , 2003

  25. [33]

    Plambeck, B

    E. Plambeck, B. Fu, S. Robinson, and R. Suri. Sample-pat h optimization of convex stochastic performance functions. Math. Program., 75(2), pp. 137–176, 1996

  26. [34]

    M. Putinar. Positive polynomials on compact semi-alge braic sets, Ind. Univ. Math. J. 42 (1993), pp. 969–984

  27. [35]

    Robinson

    S. Robinson. Analysis of sample-path optimization. Mathematics of Operations Research , 21(3), pp. 513–528, 1996

  28. [36]

    Santoso, S

    T. Santoso, S. Ahmed, M. Goetschalckx, and A. Shapiro. A stochastic programming ap- proach for supply chain network design under uncertainty. European Journal of Operational Research, 167(1), pp. 96–115, 2005

  29. [37]

    Scheiderer

    C. Scheiderer. Positivity and sums of squares: A guide t o recent results. Emerging Appli- cations of Algebraic Geometry (M. Putinar, S. Sullivant, ed s.), IMA Volumes Math. Appl. 149, Springer, 2009, pp. 271–324

  30. [38]

    Shalev-Shwartz, O

    S. Shalev-Shwartz, O. Shamir, N. Srebro, and K. Sridhar an. Stochastic convex optimization. In COLT, 2009

  31. [39]

    Shapiro, D

    A. Shapiro, D. Dentcheva, and A. Ruszczy´ nski.Lectures on stochastic programming: modeling and theory . SIAM, 2009

  32. [40]

    Sturm, SeDuMi 1.02: a MATLAB toolbox for optimizatio n over symmetric cones, Opti- mization Methods and Software , 11&12, pp

    J. Sturm, SeDuMi 1.02: a MATLAB toolbox for optimizatio n over symmetric cones, Opti- mization Methods and Software , 11&12, pp. 625–653, 1999. http://sedumi.mcmaster.ca/

  33. [41]

    Verweij, S

    B. Verweij, S. Ahmed, A. Kleywegt, G. Nemhauser, and A. S hapiro. The sample average approximation method applied to stochastic routing proble ms: a computational study. Com- putational Optimization and Applications , 24(2-3):289–333, 2003. Jiaw ang Nie, Suhan Zhong, Departmen...

Pith tools

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