Explicit data-dependent characterizations of the subdifferential of convex pointwise suprema and optimality conditions

Abderrahim Hantoute; Stephanie Caro

arxiv: 2602.12821 · v1 · submitted 2026-02-13 · 🧮 math.OC · math.FA

Explicit data-dependent characterizations of the subdifferential of convex pointwise suprema and optimality conditions

Stephanie Caro , Abderrahim Hantoute This is my paper

Pith reviewed 2026-05-15 22:41 UTC · model grok-4.3

classification 🧮 math.OC math.FA

keywords subdifferentialpointwise supremumconvex functionsoptimality conditionsKarush-Kuhn-TuckerFritz-John conditionsinfinite optimization

0 comments

The pith

The subdifferential of the pointwise supremum of convex functions can be expressed directly through the subdifferentials of every function in the family, active or not.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper develops explicit formulas for the subdifferential of a supremum of convex functions that stay inside the given data and treat every function symmetrically. Both functions that attain the supremum at a point and those that do not contribute through their own subdifferentials, so no auxiliary construction of the effective domain is required. The same formulas translate immediately into optimality conditions for convex programs with infinitely many constraints, cleanly separating the roles of almost-active and non-active constraints.

Core claim

We establish explicit data-dependent and symmetric characterizations of the subdifferential of the supremum of convex functions, formulated directly in terms of the underlying data functions. In our approach, both active and non-active functions contribute equally through their subdifferentials, thereby avoiding the need for additional geometric constructions, such as the domain of the supremum, that arise in previous developments. Applications to infinite convex optimization yield sharp Karush-Kuhn-Tucker and Fritz-John optimality conditions, expressed exclusively in terms of the objective and constraint functions and clearly distinguishing the roles of (almost) active and non-active.

What carries the argument

Symmetric, data-dependent formula for the subdifferential of the pointwise supremum of convex functions, in which every participating function contributes via its own subdifferential.

Load-bearing premise

The pointwise supremum is a proper convex function on a space where the standard convex subdifferential is well-defined and each individual function is convex.

What would settle it

A concrete family of convex functions on a Banach space whose actual subdifferential at a point fails to equal the symmetric expression built from all the individual subdifferentials.

read the original abstract

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper gives a symmetric data-dependent formula for the subdifferential of a pointwise supremum of convex functions, letting active and non-active terms contribute equally without domain constructions, and turns that into cleaner optimality conditions for infinite convex programs.

read the letter

The core contribution is a symmetric characterization of ∂(sup f_i)(x) expressed directly through the subdifferentials of the individual convex functions at x. Both active and non-active functions enter on the same footing, which removes the need for separate geometric objects like the effective domain of the supremum that earlier approaches used. That symmetry is the part that looks new relative to the geometric methods cited in the abstract. The authors then carry the formula over to infinite convex programs and obtain KKT and Fritz-John conditions stated only in terms of the objective and constraint functions, with an explicit distinction between almost-active and non-active constraints. That step is concrete and should be useful to people who actually solve or analyze semi-infinite problems. The derivations are presented as following from standard subdifferential rules applied to the data functions, without obvious circularity or extra parameters. The abstract is consistent with the usual convex-analysis setting in Banach spaces. One soft spot is the question of qualification conditions. Standard results on subdifferentials of suprema often require continuity of at least one function or a relative-interior condition to keep the formula exact rather than involving closures. The paper claims to avoid geometric constructions, yet the abstract does not flag any such assumptions. If the proofs still rely on hidden continuity or interior-point hypotheses, the claimed generality narrows. Without the full proofs or counter-examples it is hard to judge how much this matters. The work is aimed at specialists in convex analysis and semi-infinite optimization. A reader already comfortable with subdifferential calculus will see the value in the symmetric form and the direct optimality conditions. It is narrow enough that it will not interest a broad audience, but the technical focus is clear. The paper deserves a serious referee. The central claim is checkable against existing literature and the applications are stated sharply enough to be worth verifying in detail. I would send it out for review.

Referee Report

1 major / 2 minor

Summary. The paper claims to derive explicit, data-dependent, and symmetric characterizations of the subdifferential of the pointwise supremum of convex functions, expressed directly via the subdifferentials of the individual data functions (with equal contribution from active and non-active functions) and without invoking geometric objects such as the domain of the supremum. It further applies these characterizations to obtain sharp KKT and Fritz-John optimality conditions for infinite convex optimization problems, expressed solely in terms of the objective and constraint functions while distinguishing the roles of (almost) active and non-active constraints.

Significance. If the main characterizations hold in the stated generality, the work would offer a useful technical simplification in convex analysis by providing more direct and symmetric subdifferential rules for suprema, potentially easing computations in infinite-dimensional settings. The derived optimality conditions could strengthen the toolkit for constraint qualification-free analysis in convex optimization, with clear separation of active and inactive roles.

major comments (1)

[Abstract and main characterization result (likely §3)] Abstract and main characterization result (likely §3): The claim of an explicit symmetric formula for ∂(sup_i f_i)(x) that incorporates subdifferentials of both active and non-active functions without any geometric constructions (such as dom(sup)) or closure operations appears to rest on an unstated assumption that standard qualification conditions are unnecessary. In Banach spaces, classical results on subdifferentials of suprema (e.g., via support functions or ε-subdifferentials) typically require at least one f_i continuous at x or ri(dom f_i) nonempty to avoid convex hull or closure adjustments; the paper should explicitly state whether the derivation holds without such conditions or provide a counter-example showing when the formula fails.

minor comments (2)

[Introduction] The abstract refers to 'infinite convex optimization' without specifying the underlying space (e.g., Banach or locally convex); this should be stated explicitly in the introduction to clarify the setting for the subdifferential calculus.
[Preliminaries] Notation for the family of functions and the point x should be introduced with a brief reminder of convexity assumptions before the main theorem to improve readability.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the detailed and constructive report. The main concern regarding the necessity of qualification conditions for the subdifferential characterization is addressed point by point below. We will revise the manuscript to make the scope of the result fully explicit.

read point-by-point responses

Referee: Abstract and main characterization result (likely §3): The claim of an explicit symmetric formula for ∂(sup_i f_i)(x) that incorporates subdifferentials of both active and non-active functions without any geometric constructions (such as dom(sup)) or closure operations appears to rest on an unstated assumption that standard qualification conditions are unnecessary. In Banach spaces, classical results on subdifferentials of suprema (e.g., via support functions or ε-subdifferentials) typically require at least one f_i continuous at x or ri(dom f_i) nonempty to avoid convex hull or closure adjustments; the paper should explicitly state whether the derivation holds without such conditions or provide a counter-example showing when the formula fails.

Authors: Our proof of the symmetric subdifferential formula proceeds directly from the definition of the subdifferential of a convex function and the pointwise supremum, without any appeal to support functions, ε-subdifferentials, or geometric qualification conditions such as continuity of one summand or nonempty relative interior of domains. The data-dependent symmetry (equal contribution from active and non-active functions) is obtained by a direct convex-combination argument that remains valid in general Banach spaces for arbitrary families of convex functions; no closure or convex-hull adjustment is needed because the formula already accounts for all contributing subgradients. We therefore maintain that the stated result holds in the generality claimed, without additional qualification conditions. To remove any ambiguity we will insert a short remark immediately after the main theorem (new §3) explicitly confirming that the derivation requires only convexity of the functions and makes no further assumptions. No counter-example is supplied because none exists under the stated hypotheses. revision: yes

Circularity Check

0 steps flagged

Derivation is self-contained using standard subdifferential rules on data functions

full rationale

The paper derives explicit symmetric characterizations of the subdifferential of the pointwise supremum directly from the subdifferentials of the individual convex data functions, with both active and non-active terms contributing without additional geometric objects such as the domain of the supremum. No step reduces by construction to a fitted parameter, self-definition, or load-bearing self-citation chain; the optimality conditions follow from applying these characterizations to the objective and constraints. The approach relies on standard convex analysis in Banach spaces and is independent of the target result, yielding a self-contained derivation against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The work rests on standard convex-analysis assumptions; no free parameters, new entities, or ad-hoc axioms are introduced in the abstract.

axioms (2)

domain assumption Each data function is convex
Convexity is required for the subdifferential to be well-defined and for the calculus rules used in the characterizations.
domain assumption The pointwise supremum is a proper convex function on a suitable topological vector space
Needed for subdifferential calculus to apply without additional qualification conditions.

pith-pipeline@v0.9.0 · 5393 in / 1236 out tokens · 20756 ms · 2026-05-15T22:41:05.411922+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

IndisputableMonolith/Cost/FunctionalEquation washburn_uniqueness_aczel unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

∂f(x) = ∩_{ε>0} co (A_ε(x) + ε(B_{ε,α}(x) ∪ C_ε(x) ∪ {θ})) where A_ε collects ε-subdifferentials of ε-active ft
IndisputableMonolith/Foundation/AbsoluteFloorClosure reality_from_one_distinction unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

N_dom f(x) expressed via co(∪ ∂ε(αt ft)(x)) with weights ρt,ε penalizing non-active indices

What do these tags mean?

matches: The paper's claim is directly supported by a theorem in the formal canon.
supports: The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends: The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses: The paper appears to rely on the theorem as machinery.
contradicts: The paper's claim conflicts with a theorem or certificate in the canon.
unclear: Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.