Tail Bounds via Southwest Boundary
Pith reviewed 2026-05-09 22:48 UTC · model grok-4.3
The pith
Tail probabilities P(g(X) ≥ t) are bounded by n times the southwest-boundary intersection parameter s_t under continuity and monotonicity assumptions on g.
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
Core claim
Using the southwest boundary, the authors establish that P(g(X) ≥ t) ≤ n s_t, with s_t the intersection point. For homogeneous g of degree k plus constant, with identical tails f, the bound is P ≤ n f( ((t-C)^{1/k}) / (sum |a_i|)^{1/k} ). They apply this to bound the trace of a Schur multiplier on random matrices.
What carries the argument
The southwest boundary ∂_SW Q(g^{-1}[t,∞)) of the reflected superlevel set, which determines s_t by its intersection with the line L(s) of inverse tail quantiles.
Load-bearing premise
The natural continuity, symmetry, and monotonicity assumptions on g and the well-definedness of the southwest boundary and its intersection properties.
What would settle it
A specific g and set of marginal tail functions f_i satisfying the assumptions for which the computed probability P(g(X) ≥ t) is larger than n s_t at some t > 0.
read the original abstract
We derive upper bounds for probabilities of the form $P(g(\mathbf{X})\geq t)$ using the southwest boundary (recently introduced in our previous work) $\partial_{\mathrm{SW}} Q(g^{-1}[t,\infty))$, where $Q$ is a reflection to the first quadrant. Under natural continuity, symmetry, and monotonicity assumptions on $g$, this yields explicit and computable bounds of the form $P(g(\mathbf{X})\ge t)\le ns_t$, where $s_t$ is the unique parameter at which the line $L(s)=(f_1^{-1}(s),\dots,f_n^{-1}(s))$ intersects the southwest boundary. In particular, when $g$ is a homogeneous polynomial of degree $k$ (plus a constant $C$) and all tail bounds on the random variables are identical, the bound proves to the closed-form expression $$ P(g(\mathbf{X})\ge t)\leq nf\bigg(\frac{(t-C)^{1/k}}{(\sum_i|a_i|)^{1/k}}\bigg) $$ where $a_i$ are the coefficients of the monomials in $g$. We then obtain an explicit tail bound for the trace of a Schur multiplier acting on random matrices with identical tail bounds on the random variables. No assumptions are made about independence or dependence.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper derives upper bounds on tail probabilities P(g(X) ≥ t) for a function g of random variables X_i with given marginal tail bounds f_i, by locating the intersection parameter s_t of the equal-quantile line L(s) = (f_1^{-1}(s), …, f_n^{-1}(s)) with the southwest boundary ∂_SW Q(g^{-1}[t, ∞)) of the reflected sublevel set. Under continuity, symmetry, and monotonicity assumptions on g, this produces the explicit bound P(g(X) ≥ t) ≤ n s_t. For homogeneous polynomials of degree k plus constant C with identical marginal tails, the bound simplifies to the closed form P ≤ n f( (t−C)^{1/k} / (∑ |a_i|)^{1/k} ). An application to the trace of a Schur multiplier on random matrices with identical marginal tails is given, with no independence or dependence assumptions required.
Significance. If the geometric construction and containment argument hold, the approach supplies a dependence-robust, explicitly computable tail bound that avoids union-bound looseness in the equal-quantile direction and yields closed forms for polynomial g. The Schur-multiplier application illustrates utility in matrix concentration settings. The reliance on the southwest boundary from prior work is a strength when the intersection uniqueness and monotonicity implications are fully verified.
major comments (2)
- The central containment argument—that {g(X) ≥ t} is contained in the union of the marginal tail events when L(s_t) lies on the southwest boundary—relies on the monotonicity assumption on g, but the manuscript does not explicitly verify that this monotonicity together with the definition of ∂_SW Q(g^{-1}[t, ∞)) guarantees the event inclusion for arbitrary dependence structures.
- Uniqueness of the intersection parameter s_t is asserted under the listed continuity/symmetry/monotonicity assumptions on g, yet no proof or reference to a prior result establishing that the line L(s) intersects the southwest boundary at exactly one point is provided in the main theorem or its proof.
minor comments (2)
- The reflection operator Q is used without an explicit coordinate-wise definition or diagram in the introduction; adding one would clarify how the southwest boundary is constructed from the sublevel set.
- In the homogeneous-polynomial closed form, the function f is the common marginal tail bound, but this should be stated explicitly when the expression is first displayed.
Simulated Author's Rebuttal
We thank the referee for their thorough review and valuable feedback on our manuscript. We are pleased that the referee recognizes the potential of the southwest boundary approach for obtaining dependence-robust tail bounds. Below, we address each major comment in detail.
read point-by-point responses
-
Referee: The central containment argument—that {g(X) ≥ t} is contained in the union of the marginal tail events when L(s_t) lies on the southwest boundary—relies on the monotonicity assumption on g, but the manuscript does not explicitly verify that this monotonicity together with the definition of ∂_SW Q(g^{-1}[t, ∞)) guarantees the event inclusion for arbitrary dependence structures.
Authors: We acknowledge that the containment argument, while central, is not spelled out in full detail in the current version. The monotonicity of g ensures that the sublevel set g^{-1}[t, ∞) is 'upward closed' in the appropriate sense, so that its reflection Q(g^{-1}[t, ∞)) has a southwest boundary that separates the origin from the set. Consequently, any point x with g(x) ≥ t must have at least one coordinate x_i ≥ f_i^{-1}(s_t) when the equal-quantile line intersects at s_t. This inclusion holds irrespective of the joint distribution of the X_i, as it is a deterministic geometric fact about the set. We will add an explicit lemma (Lemma 2.3 in the revised manuscript) that proves {x : g(x) ≥ t} ⊆ ∪_i {x_i ≥ f_i^{-1}(s_t)} under the stated assumptions on g. This will clarify that the subsequent union bound applies for any dependence structure. revision: yes
-
Referee: Uniqueness of the intersection parameter s_t is asserted under the listed continuity/symmetry/monotonicity assumptions on g, yet no proof or reference to a prior result establishing that the line L(s) intersects the southwest boundary at exactly one point is provided in the main theorem or its proof.
Authors: The uniqueness of s_t follows from the continuity of g and the monotonicity properties that make the southwest boundary a continuous curve starting from the axes and extending to infinity in the first quadrant. Specifically, define h(s) as the minimal λ such that λ L(s) intersects the boundary; under the assumptions, h(s) is continuous and strictly increasing from 0 to ∞ as s goes from 0 to 1, guaranteeing exactly one s where h(s)=1. While this is implicit in the geometric construction from our prior work on the southwest boundary, we agree that an explicit argument is needed. In the revision, we will include a short proof of uniqueness as Proposition 2.2, relying only on the continuity and monotonicity assumptions listed in the paper. revision: yes
Circularity Check
No significant circularity in the derivation chain
full rationale
The paper locates s_t as the intersection of the equal-quantile line L(s) with the southwest boundary of the reflected sublevel set {g ≥ t}, then invokes the union bound to obtain P(g(X) ≥ t) ≤ n s_t. This union bound holds for arbitrary dependence and does not depend on any fitted parameter or re-expression of the target probability. In the homogeneous-polynomial case the closed form follows by direct diagonal evaluation g(u,…,u) = C + (∑|a_i|)u^k, solving for the level-set value of u without additional assumptions or self-referential steps. The reference to prior work supplies only the definition and properties of the boundary; the containment argument under monotonicity and the final bound are derived independently in the present manuscript and do not reduce to the inputs by construction.
Axiom & Free-Parameter Ledger
axioms (2)
- domain assumption g is continuous, symmetric, and monotone
- domain assumption The southwest boundary ∂_SW Q(g^{-1}[t,∞)) is well-defined and has the intersection properties stated in the prior work
Reference graph
Works this paper leans on
-
[1]
On Sharpest Tail Bounds for Functions of Tail Bounded Random Variables
Stephen Jordan Harrison.On Sharpest Tail Bounds for Functions of Tail Bounded Random Variables. arXiv:2604.04267 [math.PR] (version 2), April 2026. PhD thesis, University of New Mexico, 2025 (77 pages)
work page internal anchor Pith review Pith/arXiv arXiv 2026
-
[2]
Two-sided bounds for the tracial seminorm of multilinear schur multipli- ers,
A. Skripka, “Two-sided bounds for the tracial seminorm of multilinear schur multipli- ers,”Linear Algebra and its Applications, vol. 700, pp. 158–183, 2024
work page 2024
-
[3]
Vershynin,High-Dimensional Probability: An Introduction with Applications in Data Science
R. Vershynin,High-Dimensional Probability: An Introduction with Applications in Data Science. Cambridge: Cambridge University Press, 2018
work page 2018
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.