arxiv: 2605.12202 · v2 · submitted 2026-05-12 · 🧮 math.NA · cs.NA· math.OC

Recognition: no theorem link

Cavity shape reconstruction with a homogeneous Robin condition via a constrained coupled complex boundary method with ADMM

El Mehdi Cherrat, Julius Fergy Tiongson Rabago, Lekbir Afraites, Mustapha Essahraoui

Pith reviewed 2026-05-13 03:38 UTC · model grok-4.3

classification 🧮 math.NA cs.NAmath.OC

keywords shape reconstructioninverse problemRobin conditioncoupled complex boundary methodADMMshape optimizationCauchy datafinite element method

0 comments

The pith

An unknown Robin boundary segment is recovered by minimizing the imaginary part of a complex solution under inequality constraints solved via ADMM.

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

The paper addresses the inverse problem of identifying an inaccessible boundary portion subject to a homogeneous Robin condition from Cauchy data measured on the accessible boundary. A single measurement can correspond to infinitely many admissible domains, so the overdetermined problem is recast as one complex boundary value problem whose Robin condition couples the Cauchy data. The authors minimize a cost functional built from the imaginary part of the complex solution, add inequality constraints drawn from prior bounds on the state for stability, and solve the resulting shape optimization problem with ADMM in a finite-element setting. This matters because it supplies a numerically stable reconstruction procedure that remains effective under noise and varied initializations.

Core claim

The unknown boundary is reconstructed by minimizing a cost functional formed from the imaginary part of the solution to the complex boundary value problem with the coupled complex Robin condition, subject to inequality constraints on the state, with the minimization performed by the alternating direction method of multipliers (ADMM) after derivation of the shape derivative and gradient.

What carries the argument

The coupled complex boundary method that converts the Cauchy problem into a single complex BVP whose Robin condition links the data, augmented by ADMM to enforce the inequality constraints during shape optimization.

If this is right

Derivation of the shape derivative of the complex state supplies the gradient needed for efficient gradient-based shape optimization.
The added inequality constraints improve stability against noisy data and poor initial guesses compared with unconstrained formulations.
Finite-element implementation of the ADMM scheme produces usable reconstructions for several representative cavity geometries.
The approach extends existing shape-optimization techniques to the homogeneous Robin case while retaining the ability to handle non-uniqueness through optimization.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

The ADMM splitting structure could accommodate additional regularization terms or parallel computation for larger domains without major redesign.
The same complexification-plus-constraint pattern may transfer to related inverse geometric problems such as those arising in electrical impedance tomography or nondestructive testing.
Systematic variation of the admissible bounds would quantify how sensitive the final shape is to the choice of prior information.
Extension to time-dependent or nonlinear forward models would test whether the core reformulation remains effective beyond the steady elliptic setting.

Load-bearing premise

The complex Robin reformulation faithfully represents the original inverse problem, and the chosen prior bounds produce useful inequality constraints without introducing systematic bias.

What would settle it

Numerical tests on synthetic data for a known exact cavity in which the recovered boundary deviates substantially from the true shape even in the noise-free case, or in which the iteration fails to converge for moderate noise, would show the method does not achieve reliable reconstruction.

Figures

Figures reproduced from arXiv: 2605.12202 by El Mehdi Cherrat, Julius Fergy Tiongson Rabago, Lekbir Afraites, Mustapha Essahraoui.

**Figure 2.** Figure 2: Comparison of results for different regularization parameters ( [PITH_FULL_IMAGE:figures/full_fig_p012_2.png] view at source ↗

**Figure 3.** Figure 3: Cost function and gradient norm evolution for the Ellipse (E) and L-block (L) cases. [PITH_FULL_IMAGE:figures/full_fig_p012_3.png] view at source ↗

**Figure 4.** Figure 4: Real and imaginary parts of the state solution at [PITH_FULL_IMAGE:figures/full_fig_p012_4.png] view at source ↗

**Figure 5.** Figure 5: Evolution of the real and imaginary parts of the state solution for [PITH_FULL_IMAGE:figures/full_fig_p013_5.png] view at source ↗

**Figure 6.** Figure 6: Evolution of the real and imaginary parts of the adjoint solution for [PITH_FULL_IMAGE:figures/full_fig_p014_6.png] view at source ↗

**Figure 7.** Figure 7: Comparison of the ADMM-based reconstruction using ( [PITH_FULL_IMAGE:figures/full_fig_p020_7.png] view at source ↗

**Figure 8.** Figure 8: Corresponding histories of values for the cost, gradient norm, and Hausdorff distances [PITH_FULL_IMAGE:figures/full_fig_p020_8.png] view at source ↗

**Figure 9.** Figure 9: Comparison of the ADMM-based reconstruction using ( [PITH_FULL_IMAGE:figures/full_fig_p020_9.png] view at source ↗

**Figure 10.** Figure 10: Corresponding histories of values for the cost, gradient norm, and Hausdorff dis [PITH_FULL_IMAGE:figures/full_fig_p021_10.png] view at source ↗

**Figure 11.** Figure 11: Effect of size and location of the initial guess with [PITH_FULL_IMAGE:figures/full_fig_p021_11.png] view at source ↗

**Figure 12.** Figure 12: Reconstructions with varying ρ = 2, 5, 10 under exact measurements. 21 [PITH_FULL_IMAGE:figures/full_fig_p021_12.png] view at source ↗

**Figure 13.** Figure 13: Reconstructions under noisy measurements with low noise levels [PITH_FULL_IMAGE:figures/full_fig_p022_13.png] view at source ↗

**Figure 14.** Figure 14: Reconstructions under noisy measurements with low noise levels [PITH_FULL_IMAGE:figures/full_fig_p023_14.png] view at source ↗

**Figure 15.** Figure 15: Reconstructions under noisy measurements with noise level [PITH_FULL_IMAGE:figures/full_fig_p024_15.png] view at source ↗

**Figure 16.** Figure 16: Shape evolution of two of the reconstructed shapes in Figure [PITH_FULL_IMAGE:figures/full_fig_p025_16.png] view at source ↗

**Figure 17.** Figure 17: Reconstructions with multiple concave regions under different noise levels [PITH_FULL_IMAGE:figures/full_fig_p025_17.png] view at source ↗

read the original abstract

We revisit the problem of identifying an unknown portion of a boundary subject to a Robin condition based on a pair of Cauchy data on the accessible part of the boundary. It is known that a single measurement may correspond to infinitely many admissible domains. Nonetheless, numerical strategies based on shape optimization have been shown to yield reasonable reconstructions of the unknown boundary. In this study, we propose a new application of the coupled complex boundary method to address this class of inverse boundary identification problems. The overdetermined problem is reformulated as a complex boundary value problem with a complex Robin condition that couples the Cauchy data on the accessible boundary. The reconstruction is achieved by minimizing a cost functional constructed from the imaginary part of the complex-valued solution. To improve stability with respect to noisy data and initialization, we augment the formulation with inequality constraints through prior admissible bounds on the state, leading to a constrained shape optimization problem. The shape derivative of the complex state and the corresponding shape gradient of the cost functional are derived, and the resulting problem is solved using an alternating direction method of multipliers (ADMM) framework. The proposed approach is implemented using the finite element method and validated through various numerical experiments.

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 adds state constraints and an ADMM solver to the coupled complex boundary method for Robin cavity reconstruction, and the numerics look workable but thin on quantitative checks.

read the letter

The main takeaway is that this work takes the coupled complex boundary reformulation for an inverse Robin boundary problem, adds inequality constraints on the state for stability, and solves the resulting shape optimization with ADMM plus finite elements. The authors derive the shape derivative of the complex state and the gradient of the cost functional built from the imaginary part, then run several test cases with noisy data that produce visually reasonable cavity shapes. Non-uniqueness of the inverse problem is noted up front, which is honest. The implementation follows standard FEM practices and the ADMM splitting is applied in a straightforward way to the constrained problem. That combination is presented as a new packaging for this class of problems, and the derivations check out on the surface without obvious algebraic errors. The experiments demonstrate that the method can handle some noise and different initial guesses, which is the practical point. The soft spots sit in the validation. Results are shown mostly through plots without tables of L2 errors, convergence rates, or side-by-side comparisons against the unconstrained version or other shape optimization schemes like level-set or gradient descent methods. The choice of admissible bounds for the constraints is described at a high level but not tested for sensitivity, so it is hard to judge how much prior information is doing the heavy lifting. Parameter choices for ADMM are not explored in depth either. This is a targeted numerical methods paper for people already working on inverse boundary identification or shape optimization in PDEs. A reader who needs a stable solver for similar overdetermined Robin problems could extract the ADMM setup or the constraint idea. It is incremental rather than foundational, but the framework is coherent and the code path is clear enough to reproduce. I would send it for peer review. The core math and discretization are solid, and referees can push for more quantitative benchmarks and sensitivity tests without rewriting the contribution.

Referee Report

3 major / 2 minor

Summary. The paper proposes a constrained version of the coupled complex boundary method for reconstructing an unknown cavity boundary subject to a homogeneous Robin condition, given Cauchy data on the accessible boundary. The overdetermined inverse problem is recast as a complex BVP that couples the data via a complex Robin condition; a cost functional is then minimized whose integrand is the imaginary part of the complex state. Inequality constraints based on prior admissible bounds on the state are added for stability, the shape derivative and gradient are derived, and the resulting constrained shape optimization problem is solved by an ADMM scheme discretized with finite elements. Numerical experiments are presented to illustrate the approach.

Significance. If the numerical behavior holds under quantitative scrutiny, the work supplies a practical, stabilized algorithm for a classically non-unique inverse boundary problem. The combination of complex coupling, explicit inequality constraints, and ADMM is a coherent extension of existing shape-optimization techniques and could be useful for other ill-posed cavity or inclusion identification tasks in which single-measurement data are available.

major comments (3)

Abstract and §4 (numerical experiments): the claim of improved stability with respect to noisy data and initialization is not supported by quantitative error tables, convergence rates, or systematic comparisons against unconstrained or alternative methods (e.g., standard shape-gradient descent or level-set approaches). Without such metrics it is impossible to judge whether the added constraints and ADMM actually deliver the advertised robustness.
§3.2 (inequality constraints): the admissible bounds on the state that define the inequality constraints are introduced without a priori justification or sensitivity analysis. It is therefore unclear whether these bounds can be chosen in a data-driven way or whether they introduce systematic bias that affects the reconstructed cavity shape for realistic noise levels.
§2.1 (complex Robin reformulation): while the coupling of Cauchy data into a single complex Robin condition is formally correct, the manuscript does not verify that the imaginary-part cost functional remains equivalent to the original least-squares misfit when the Robin coefficient is homogeneous; a short consistency check or reference to the earlier coupled-complex literature would strengthen the derivation.

minor comments (2)

Notation: the distinction between the physical Robin coefficient (zero) and the complex coupling parameter should be made explicit in the first appearance of the complex BVP.
Figure captions: several numerical figures lack axis labels or color-bar scales, making it difficult to read the reported state values or residual magnitudes.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive comments on our manuscript. We have revised the paper to address the concerns raised, particularly by enhancing the numerical validation and providing additional justifications and checks. Our point-by-point responses are as follows.

read point-by-point responses

Referee: Abstract and §4 (numerical experiments): the claim of improved stability with respect to noisy data and initialization is not supported by quantitative error tables, convergence rates, or systematic comparisons against unconstrained or alternative methods (e.g., standard shape-gradient descent or level-set approaches). Without such metrics it is impossible to judge whether the added constraints and ADMM actually deliver the advertised robustness.

Authors: We agree that quantitative support is essential for validating the stability claims. The original manuscript presented numerical results primarily through figures showing reconstructions under noise and different initializations, with qualitative discussions of robustness. To strengthen this, we have added a new subsection in §4 with error tables reporting L2 norms of the shape error for varying noise levels (0%, 1%, 5%, 10%) and different initial guesses. We also include comparisons with the unconstrained coupled complex method and a standard gradient descent approach, showing lower errors and better convergence with the constrained ADMM. These additions provide the requested metrics and confirm the improved stability. revision: yes
Referee: §3.2 (inequality constraints): the admissible bounds on the state that define the inequality constraints are introduced without a priori justification or sensitivity analysis. It is therefore unclear whether these bounds can be chosen in a data-driven way or whether they introduce systematic bias that affects the reconstructed cavity shape for realistic noise levels.

Authors: The bounds are selected based on a priori estimates derived from the maximum principle applied to the elliptic problem and the known range of the Robin coefficient, ensuring they are consistent with the physical setting. We acknowledge the lack of sensitivity analysis in the original submission. In the revision, we have added a paragraph in §3.2 discussing the choice of bounds and included a sensitivity study in the numerical section, demonstrating that the reconstructions remain stable for bounds varied by ±20% around the nominal values, with no significant bias introduced at the noise levels tested. We also outline a data-driven approach using an initial unconstrained solve to estimate the state range. revision: yes
Referee: §2.1 (complex Robin reformulation): while the coupling of Cauchy data into a single complex Robin condition is formally correct, the manuscript does not verify that the imaginary-part cost functional remains equivalent to the original least-squares misfit when the Robin coefficient is homogeneous; a short consistency check or reference to the earlier coupled-complex literature would strengthen the derivation.

Authors: We appreciate this point. In the revised manuscript, we have inserted a brief consistency verification in §2.1. Specifically, we show that under the homogeneous Robin condition, the imaginary part of the complex solution on the accessible boundary coincides with the difference between the two Cauchy data components, making the cost functional equivalent (up to a multiplicative constant) to the standard least-squares misfit. This is derived by separating real and imaginary parts of the complex Robin condition. We have also added references to the foundational papers on the coupled complex boundary method to contextualize the approach. revision: yes

Circularity Check

0 steps flagged

No significant circularity

full rationale

The derivation begins with an explicit reformulation of the overdetermined Cauchy problem into a complex BVP with a coupled Robin condition that directly incorporates the given data pair. A cost functional is then defined directly from the imaginary part of the resulting complex state, shape derivatives are computed from first principles, inequality constraints are added from prior admissible bounds, and the problem is solved via ADMM. None of these steps reduces a claimed prediction or result to a fitted parameter, self-referential quantity, or unverified self-citation by construction. The paper acknowledges non-uniqueness and presents the method as producing reasonable numerical reconstructions rather than exact recovery. No load-bearing ansatz, uniqueness theorem, or renaming of known results is invoked in a way that collapses the chain to its inputs.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on the reformulation of the inverse problem as a complex BVP and the validity of the chosen cost functional and constraints; these are standard domain assumptions in inverse PDE problems but are not independently verified in the abstract.

axioms (1)

domain assumption The overdetermined Cauchy data problem can be equivalently reformulated as a complex boundary value problem with a complex Robin condition that couples the measurements.
This is the foundational step stated in the abstract for converting the inverse problem into an optimization task.

pith-pipeline@v0.9.0 · 5532 in / 1355 out tokens · 164345 ms · 2026-05-13T03:38:27.924294+00:00 · methodology