Papers

For the Artin group of type $D_5$, we determine its automorphism group and the automorphism group of its quotient by the center. This settles the only remaining case, $n=5$, in the classification of automorphisms of Artin groups of spherical type $D_n$.

Lyapunov-style convergence proofs, which establish a nonincreasing sequence to provide a quantitative convergence rate for an algorithm, are popular and often considered desirable in first-order optimization. However, existing approaches to finding such Lyapunov functions rely on hand-designed templates or prior insight on the proof structure, and do not certify that the resulting Lyapunov-style analysis provides the sharpest convergence bound. In this work, we introduce a systematic framework for converting a tight, analytic convergence proof of an optimization algorithm, often found via computer assistance, into an equivalent proof based on Lyapunov functions. We implement a concrete procedure that combines a performance estimation problem (PEP) toolbox with elementary linear algebra, and show that it captures a number of prior Lyapunov analyses within a single Jupyter notebook. Based on our implementation, a user can straightforwardly test our systematic and interactive procedure on their own optimization algorithm of interest to search for a tight Lyapunov-style proof via code, without the need to comprehend the implementation details. We extend the application of our framework and discover four novel analytic Lyapunov-style proofs, where notably, one of them identifies a new exact optimal proximal algorithm for strongly monotone inclusion problems.

Synthetic data generation is a powerful approach to privacy-preserving statistical analysis, where data-release mechanisms are governed by a privacy-utility tradeoff: they should provide privacy guarantees while preserving the statistical utility of confidential data. We develop a Bayesian nonparametric framework for private synthetic data generation tailored to discrete data. Specifically, the confidential data are modeled as a random sample from an unknown discrete distribution endowed with a Pitman-Yor process prior, and synthetic data are generated from the corresponding posterior-predictive distribution. Since the Pitman-Yor process defines an almost surely discrete random probability measure, the resulting mechanism is naturally suited to data with ties and settings involving a potentially large, unknown, or growing number of categories. We study differential privacy guarantees of the Pitman-Yor posterior-predictive mechanism across the three regimes of the discount parameter $\sigma\in(-\infty,1)$. For $\sigma\in(0,1)$, we establish an instance-level $(\varepsilon,\delta)$-differential privacy guarantee. For $\sigma=0$ and $\sigma<0$, corresponding respectively to the Dirichlet process prior and to a parametric Dirichlet-Multinomial model, stronger guarantees are obtained, under suitable conditions on the released sample size. We also investigate statistical utility, or informativity, of the released data via the expected $1$-Wasserstein distance between the empirical distribution of the synthetic data and the "true" data-generating distribution. For $\sigma<0$ and $\sigma=0$, we prove consistency of the empirical distribution in this metric and derive explicit convergence rates, making precise the privacy-utility tradeoff: stronger privacy guarantees impose more restrictive choices of the released sample size, slowing down convergence to the "true" data-generating distribution.

For a general curve in positive characteristic, taking the Frobenius pullback induces a generically finite rational map V on the moduli space of rank 2 vector bundles with trivial determinant. Recently, Kondo–Wakabayashi show that the generic degree of V, considered as a function on the characteristic of the base field, is a quasi-polynomial. In this paper, we show that this quasi-polynomial is indeed a polynomial, and we write out this polynomial explicitly.

We prove a localized continuous-frequency operator estimate for centered Gaussian chaoses of order two. The result applies to operator-valued centered second chaoses, including Wick-centered same-family variants, between Hilbert spaces. In the model, two Gaussian frequency legs at scale $N$, an input leg at scale $Q$, and an output leg at scale $M$ are coupled through a soft incidence kernel; non-orthogonal Gaussian profiles are represented by covariance synthesis maps. The proof combines four oriented flattenings, rectangular non-commutative Khintchine inequalities, soft-incidence Schatten bounds, and Sobolev–Besov dyadic summation. The time lift gives $L^p$ operator convergence, while a Galerkin stabilization hypothesis gives pathwise full-cutoff convergence by the first Borel–Cantelli lemma. Under $\mathcal G(N)\lesssim N^{-\Gamma}$ one obtains the window \[ \Gamma>\frac d2, \qquad s<\lambda+\Gamma-d, \qquad \max\{0,d-\Gamma\}<\sigma<\lambda+\Gamma-d. \] The theorem applies to the near-output Wick-centered branch of localized paracontrolled resonant products on $\mathbb R^d$.

We introduce a Dirichlet–multinomial (DM) deviance residualization for sparse, jointly overdispersed count matrices, the regime that dominates sequencing-based biochemical assays. The DM null treats each sample's count vector as a fixed-total composition with a single scalar concentration $\alpha_0$ governing overdispersion, and arises exactly by conditioning independent negative-binomial feature counts on the observed sample total – making the DM the joint conditional analogue of standard feature-wise overdispersed count models. The resulting transform preserves exact sparsity, evaluates in constant time per nonzero entry, agrees with multinomial residuals on singleton counts, shrinks repeated-count residuals according to the overdispersion the null tolerates, and recovers the multinomial residual as $\alpha_0\to\infty$. The same fixed-dispersion comparison principle extends to ordered and tree-structured features via the generalized DM and the Dirichlet-tree multinomial, giving a single residual family that subsumes joint and feature-wise count nulls under a common compositional logic and is computationally lightweight enough to drop into existing sparse pipelines.

For every dimension $d > 1$, we establish explicit criteria on lattices $\Lambda \subset \mathbb{R}^{2d}$ with density $D(\Lambda) > 1$ such that no function with a continuous Zak transform generates a Gabor frame along $\Lambda$. In particular, this gives a negative answer to the existence problem of Gabor frames with window functions in the Schwartz space, the Feichtinger algebra, and the Fourier-invariant Wiener space. Our result is based on a characterization of when a collection of quasiperiodic functions admits a common zero, which may be of independent interest. We also include a formalization of our main result in Lean 4.

We study the relation between lower ball growth, rearrangement oscillations, and Sobolev-type embeddings on metric measure spaces. We prove that a lower estimate for the ball function \[ h_{\mu}(r):=\inf_{x\in\Omega}\mu(B(x,r)) \] implies pointwise symmetrization inequalities for averaged differences and Hajłasz gradients, without doubling, Poincaré-type assumptions, or extra topological hypotheses. Conversely, the validity of either family of inequalities for suitable cutoff functions recovers the corresponding lower growth estimate. Thus, up to multiplicative constants, the lower geometry of the measure is equivalent to the pointwise oscillation principle itself. Taking norms in rearrangement-invariant spaces yields normed oscillation inequalities and Sobolev embeddings for Hajłasz and averaged Besov spaces. This separates the metric-measure input from the choice of the final function-space target. In the power-growth model, the resulting targets are identified with Lorentz spaces. We also show how Sobolev inequalities between general rearrangement-invariant spaces force lower ball estimates through the fundamental functions of the source and target spaces. The results apply to general admissible growth functions and moduli of smoothness, beyond the classical power setting.

We determine analytically for all $k\in\{0,1,\ldots,d-1\}$ the $k$-volume densities of a Poisson–Voronoi tessellation of intensity $\lambda>0$ in the $d$-dimensional hyperbolic space of constant curvature $-1$. This largely extends previous results of Isokawa in dimensions two and three. As applications, we provide closed form expressions for all face volume densities and all typical face volumes of the ideal Poisson–Voronoi tessellation (IPVT), which is the low-intensity limit as $\lambda\downarrow0$ of the hyperbolic Poisson–Voronoi tessellation. As a main tool we develop a new Blaschke–Petkantschin–type formula in hyperbolic space.

Let $K$ be a $p$-adic field. We define the notion of finite height for an étale $\mathbb{Z}_p$-local system on a smooth affinoid adic space $\mathcal{X}$ over $K$ with semistable reduction. Using analytic prismatic $F$-crystals and purity results of Du-Liu-Moon-Shimizu (arXiv:2404.19603), we prove that if an étale $\mathbb{Z}_p$-local system over $\mathcal{X}$ is of finite height then its pullback along a finite étale cover of $\mathcal{X}$ is semistable. This answers a question of Tong Liu in the relative setting.

In this paper, it is proved that any nonnegative integer can be written in the following form $$ x(x+1)/2 + y(3y+1)/2 + z(5z+1)/2, \qquad x,y,z \in \mathbb{N}. $$ This settles the conjecture recorded as OEIS A287616. All parts of the proof have been formalized in Lean 4, with the exception of two results: one externally cited theorem and one statement verified by symbolic computation. Both the natural-language proof and the Lean formalization were generated by the MechMath Agent Team developed by the authors.

We consider the problem of estimating the fidelity of an unknown quantum state to a known reference state to within additive error $\varepsilon$. We show that the sample complexity is $O(r^2/\varepsilon^2)$ with optimal $\varepsilon$-dependence when the reference state is of rank $r$, improving the previous best $O(r^2\log^2(1/\varepsilon)/\varepsilon^4)$ due to Utsumi, Nakata, Wang, and Takagi (QIP 2026). We also provide a lower bound of $\Omega(r/\varepsilon^2)$, improving the previous best $\Omega(r/\varepsilon+1/\varepsilon^2)$, with implications to quantum query complexity. Moreover, we further consider the case where the unknown state is of rank at most $r$ while the reference state can be arbitrary, for which the sample complexity is shown to be $O(r^2/\varepsilon^4)$. As an application, we present an approach to tolerant quantum state certification, generalizing the exact certification studied in Bădescu, O'Donnell, and Wright (STOC 2019).

We show that if a group $G$ acts on a contractible $CW$-complex with a contractible strong fundamental domain and the stabilizers of the action satisfy certain properties, then the homological Sigma invariants of the group are either empty or dense in the character sphere. This was known for several families of groups like right-angled Artin groups and pure symmetric automorphisms of free groups. We also exhibit some applications including the generalization of the previous fact to pure symmetric automorphisms of right-angled Artin groups.

Moment closures at odd truncation order present a fundamental difficulty: the standard Gramian closure saturates the realizability boundary, producing only weak hyperbolicity and failing to preserve Maxwellian equilibrium. We show that every odd-order closure for the one-dimensional kinetic equation decomposes into a boundary term, the Schur complement of the Hankel moment matrix, and a positive margin above it. An exact polynomial identity connects this margin to the eigenvalues of the flux Jacobian, reducing hyperbolicity to a root-splitting problem. A dimensional argument proves that no margin depending only on density, velocity, and temperature can produce a hyperbolic system for $M \geq 5$. A one-parameter family $C_{\eta,n}$, $\eta \in [0,1]$, built from normalized Schur-complement ratios, reveals that the Morin-McDonald closure is the arithmetic endpoint of this decomposition. The weighted AM-GM inequality makes the accuracy-robustness tradeoff precise: the geometric endpoint ($\eta = 0$) is 2-4% more accurate on bimodal benchmarks, while the arithmetic endpoint ($\eta = 1$, Morin-McDonald) provides the most robust hyperbolicity profile. All members share the same equilibrium Jacobian, whose spectral radius is 13% ($M = 5$) to 29% ($M = 13$) smaller than Grad's closure, allowing larger CFL time steps. A linearized entropy exists at every tested order, and for a source-compatible choice of the symmetrizer weights, the BGK source dissipates it near equilibrium. A smooth nonlinear entropy exists for $M = 3$ but does not for $M = 5$ or $M = 7$ (certified by linear programming). The closure is validated on bimodal and Mott-Smith benchmarks, where the interpolated family achieves errors 10-40x smaller than the Gramian or Grad closures, and demonstrated in free-transport Riemann problems at $M = 5, 7, 9, 11$ and BGK Riemann problems at $M = 5$ and $9$.

We define a canonical labelling function on the class of finite solvable groups so that two such groups $G$ and $H$ are isomorphic if and only if can$(G)=$can$(H)$. Specifically, can$(G)$ is a group presentation that describes a group isomorphic to $G$, and our description explains how to construct an isomorphism $G\to$can$(G)$. Our approach is motivated by O'Brien's (1993) canonical presentations for finite $p$-groups and utilises ideas from group cohomology first described by Robinson (1982) and automorphism group algorithms developed by Smith (1994), Holt (2001), and others. We also discuss a proof-of-concept implementation for the computer algebra system GAP and comment on the major bottlenecks and open research questions.

We present HiReLC, a hierarchical ensemble-reinforcement learning framework for automated joint quantization and structured pruning of deep neural networks. The framework decomposes the compression search across two levels of abstraction: low-level agents (LLAs) operate independently per block, selecting per-kernel configurations over a multi-discrete action space spanning bitwidth, pruning keep-ratio, quantization type, and granularity, while high-level agents (HLAs) coordinate global budget allocation via ensemble voting guided by Fisher Information-based sensitivity estimates. To mitigate the computational cost of policy evaluation, an iterative active learning loop interleaves surrogate-guided RL optimization with post-compression fine-tuning, using a lightweight MLP surrogate to amortize expensive evaluations and a logit-MSE proxy during cold-start. The surrogate is used for reward shaping rather than as a replacement for final post-compression evaluation. The controller is architecture-agnostic by design, with a modular layer abstraction decoupling the RL environment from the underlying network topology. Experiments across Vision Transformer and CNN benchmarks demonstrate effective parameter-storage compression ratios of 5.99 - 6.72$\times$ with a 3.83 % gain in one setting and 0.55 - 5.62 % accuracy drops elsewhere, supporting hierarchical policy decomposition and sensitivity-aware guidance as practical design choices for joint neural network compression.

There has been significant recent progress in algorithms for approximation of Nash equilibrium in large two-player zero-sum imperfect-information games and exact computation of Nash equilibrium in multiplayer strategic-form games. While counterfactual regret minimization and fictitious play are scalable to large games and have convergence guarantees in two-player zero-sum games, they do not guarantee convergence to Nash equilibrium in multiplayer games. Recently, an approach has been presented for exact computation of Nash equilibrium in multiplayer imperfect-information games that solves a quadratically constrained program based on a nonlinear complementarity problem formulation derived from the sequence-form game representation. This formulation was solved using Gurobi's nonconvex quadratic solver, which employs spatial branch-and-bound to iteratively refine variable bounds by solving convex relaxations of bilinear terms. During presolve, Gurobi introduces auxiliary variables and, in some cases, binary variables, leading to an internal mixed-integer quadratically constrained program. This approach was demonstrated to outperform prior algorithms from the Gambit software suite and quickly solve three-player Kuhn poker after removal of dominated actions; however, the algorithm was not able to solve the full version of the game within 24 hours. In this paper, we derive finite bounds on slack and multiplier variables in the nonlinear complementarity formulation. These bounds can strengthen the convex relaxations used within spatial branch-and-bound. We compare uniform corollary bounds with finer proposition-based bounds on full three-player Kuhn poker. All eight bounded configurations return numerically verified equilibria in under 17 seconds, while the unbounded formulation finds no feasible solution within 300 seconds. The finest bounds do not consistently yield the fastest computation.

In this note, we study the index of the Laplace operator on planar domains $\Omega$ with compact smooth boundary, allowing both the compact and noncompact cases, under a natural Robin boundary condition involving the geodesic curvature of $\partial\Omega$. We obtain lower bounds for the index in terms of the number of boundary components. Our approach combines conformal and spectral techniques with the topology of the domain, encoded in the space of $L^2$ harmonic vector fields tangent to the boundary. The result is motivated by index estimates for free boundary minimal surfaces and provides an intrinsic counterpart to a borderline geometric situation arising in that setting.

We show that a few physical principles – smoothness, homogeneity, isotropy, and the determinism of inertial motion – force the invariant interval governing the geometry of spacetime to reduce to a quadratic form, without presupposing the existence of light or electromagnetic phenomena. Formalizing these as axioms about an "invariant interval" function $D:\mathbb{R}^n\to\mathbb{R}$ ($n\geq 3$), we find that smoothness and homogeneity force $D$ to be homogeneous of degree $p > 0$; determinism – that an inertial worldline be uniquely fixed by its initial point and direction – makes its geodesics straight lines; and isotropy – that the isometry group act transitively on each level set, with the stabilizer of a reference direction reversing every transverse direction – forces $D(v) = C\,(v^T S v)^{p/2}$ for a nondegenerate symmetric matrix $S$ and $p > 0$, with $p = 2$ (so that $D$ is exactly quadratic) when $S$ is indefinite. Thus the only admissible invariant intervals are powers of nondegenerate quadratic forms. The signature of $S$ is otherwise free: the definite case is Euclidean geometry and the indefinite case includes both Minkowski and ultrahyperbolic geometries, the two cases distinguished by the absence or presence of a null cone.

We efficiently conflict-free color every planar graph with 4 colors. An (open-neighborhood) conflict-free coloring assigns colors to vertices in a way that every vertex v has a neighbor w such that the color of w is distinct from the colors of the other neighbors of v (i.e., the color of w is unique in the open neighborhood of v). A previous best upper bound on the conflict-free chromatic number of planar graphs was 5, and it is known that 4 colors are sometimes necessary. Deciding whether, e.g., a planar graph admits a conflict-free coloring with 3 colors is NP-complete. Our approach uses a refined variant of the classical Gallai-Edmonds decomposition and the Four Color Theorem. In fact, our result is equivalent to the Four Color Theorem.

We prove a Gaussian rigidity theorem for infinite exchangeable sequences of real-valued random variables: the joint Gaussianity of a single pair of entries already forces the entire sequence to be a Gaussian process. This settles a conjecture raised by Newman (2026). The main analytic ingredient in the proof is Hardy's uncertainty principle. We also obtain a finite-dimensional vector-valued extension.

Common first-order optimizers, such as Adam, implicitly treat each parameter block as an unstructured vector, which disregards the multilinear weight structure present in many modern machine learning models. Recent work has shown that exploiting matrix structure can improve optimization dynamics. A notable example is Muon, which performs steepest descent under the spectral norm constraint. We take the next step and introduce Tensorion, a tensor-aware optimizer that extends Muon's constrained optimization perspective from matrices to higher-order tensors. Tensorion is built around a linear minimization oracle (LMO) over a tensor norm ball. The norm is carefully chosen to balance two objectives: tightly bounding the tensor spectral norm, while still keeping the LMO tractable. This LMO becomes computable because it reduces to operations on adaptively selected unfolding matrices. Notably, when restricted to order-2 tensors (i.e., matrices), Tensorion recovers Muon exactly. Experiments on tensor-based computer vision problems suggest that Tensorion can offer improved convergence behavior and more stable gradient updates compared with Adam-based and existing tensor-aware baselines in the evaluated settings.

Routing and packing are intrinsically coupled in transport problems, requiring joint planning for cost-efficient and physically realizable solutions. We study a pickup and delivery problem with two-dimensional packing constraints (2P-PDP). Unlike vehicle routing variants where items are loaded before vehicles leave the depot and packing is validated only once, the 2P-PDP induces non-monotonic free-space evolution, substantially increasing feasibility-checking complexity. To address this bottleneck, we propose a generic dominance-based feasibility framework that is embeddable in a broad class of exact and heuristic routing algorithms. Under no-relocation constraints, inferring feasibility from a previously verified packing state requires preserving the pickup and delivery order of onboard items. To this end, we introduce an order-preserving mapping that jointly captures geometric containment and sequence compatibility, enabling dominance-based inference by embedding the new packing state into a verified reference plan. To further reduce dominance-screening overhead, we design three search rules to guide candidate exploration and tailored strategies to store, retrieve, and prioritize verified states. Computational experiments show that the proposed approach reduces feasibility-checking time by up to 42% compared to a benchmark without dominance. The improvement stems from reducing exact packing-procedure calls, shifting verification effort away from the most computationally expensive stage.

The bootstrap is a versatile method for quantifying statistical uncertainty. Among its variants, a popular approach, the studentized bootstrap, provably achieves higher-order coverage error reduction compared to other benchmarks. However, its implementation typically requires an analytical form of the standard error, or otherwise an additional layer of resampling effort which can be computationally expensive. In this paper, we introduce what we call the studentized cheap bootstrap that achieves the same higher-order coverage accuracy as the conventional studentization, but substantially thinning the computational effort in the additional resampling layer to only very few Monte Carlo replications. Intriguingly, while conventional wisdom views "studentization" as an informal link between the bootstrap and t-distribution, we provide a first recognition that this link is in fact formal, notably with a distinct insight that the degree of freedom in the t-distribution corresponds to the Monte Carlo computation effort in the additional resampling layer, rather than the data size as in traditional thinking. Moreover, our desirable higher-order coverage accuracy builds crucially on this insight, as well as explicit calculations and geometric analyses of higher-order terms in the Edgeworth and Cornish-Fisher expansions tailored to limiting t-distributions.

Numerical integration involving expensive target functions is a common bottleneck in Bayesian inference and simulation. When a cheap surrogate is available, standard approaches such as reweighting or importance sampling often suffer from high variance and inefficient use of function evaluations. We introduce Slice Monte Carlo integration (S$\ell$MC), a method that leverages a Nested Sampling-like procedure on the surrogate to partition the space into informative strata, or slices, while generating samples in the parameter space drawn from the prior within each slice. This enables stratified Monte Carlo integration of the expensive target function over the surrogate-induced partition, yielding an efficient estimate of the target integral. The surrogate level sets therefore make the induced ordering of the parameter space the relevant information for S$\ell$MC, rather than the pointwise target-to-surrogate ratios governing importance-sampling efficiency. Another key advantage of S$\ell$MC is the decoupling of slice volume estimation from the evaluation of the target, allowing the refinement of the slice-volume estimates without additional target evaluations. We investigate the properties of S$\ell$MC, demonstrate how to efficiently generate posterior samples, and assess its robustness in controlled Gaussian benchmark families with continuously tunable surrogate-target mismatch.

The paper deals with initial boundary value problem for the Magnetic Zakharov system arising in plasma physics in two-dimensional domains. It was proved the global well-posedness of this problem in some Sobolev type class.

In this paper, we investigate the regularity theory of local minimizers of multi-phase energy functionals. As a key feature of our work, instead of the classical Hölder continuity assumptions on the modulating coefficients and interaction between the growth exponents, we assume that these coefficients belong to a suitable class of Muckenhoupt weights. The presence of multiple growth phases with the degenerate or singular nature of Muckenhoupt weights poses substantial analytical difficulties that prevent a direct application of classical theory. Our approach requires a refinement of localized energy estimates and an adapted iteration scheme that exploits the reverse Hölder properties of the Muckenhoupt weights. As our main results, we establish the higher integrability, local boundedness, and Hölder continuity of local minimizers. Most notably, we prove the Harnack inequality for non-negative local minimizers, which, to the best of our knowledge, stands as the first result of its kind in the multi-phase setting involving Muckenhoupt modulating coefficients. This paper is a contribution toward a better understanding of the qualitative behavior of minimizers in non-uniformly elliptic variational problems, and offers a new framework that complements the existing literature beyond the classical Hölder continuity of modulating coefficients.

This theoretical note studies the finite axiomatizability of strict majority reasoning in finite social decision frames. Moss and Pedersen (2026) <doi: 10.48550/arXiv.2606.23853> introduce a coherence criterion that characterizes exactly when qualitative majority judgments are representable by a finitely additive measure. The question addressed here is whether that coherence criterion can be replaced, in the finite setting, by any bounded finite fragment. We prove that it cannot. For every $k\ge 1$, we construct a maximal standard frame whose shortest coherence violation has length exactly $2k+2$. Hence there is no uniform finite bound on the incoherence index of social decision frames, resolving Conjecture 5.7 stated by Moss and Pedersen (2026). The construction is geometric, in the sense that it proceeds via orthogonality and dimension in rational vector spaces, and self-contained: it isolates a symmetric family of half-sized voting blocs and extends it to a maximal frame in which every shorter balanced obstruction is excluded. Along the explicit infinite sequence of universe sizes obtained in the construction, this also establishes the middle-layer family predicted by Conjecture B.25 by Moss and Pedersen (2026). Together with the soundness and completeness theorem for the Moss-Pedersen minimal logic for strict majorities, this establishes that measurable social decision frames are not finitely axiomatizable in that language.

Dirichlet-to-Neumann (DtN) maps send boundary values of a partial differential equation (PDE) solution to its normal derivative on the boundary. Learning such maps across varying domains is important for boundary-value problems, but a black-box neural operator must model both the operator's non-smoothing principal behavior and its dependence on boundary geometry. We use the boundary integral representation of the DtN map to obtain, for smooth planar boundaries, a useful principal-part decomposition: a geometry-independent leading operator can be written as a universal Fourier multiplier, while the remaining geometry-dependent correction is smoother. We propose Principal-Part Decomposed Neural Operators (PPDNO), a hybrid analytic-neural framework that turns this decomposition into a geometry-conditioned operator learning model. PPDNO computes the principal part by FFT and trains a low-rank Deep Operator Network (DeepONet)-type architecture to approximate only the residual correction across families of boundary geometries. This design keeps the exact linear action on the boundary data, exposes the sampled boundary as an input to the model, and turns the learned target into a smoother operator family. We justify the decomposition theoretically by proving smoothing and separated-approximation properties of the residual, and we derive finite-node and training-error bounds for the reconstructed full map. Experiments on interior Laplace problems over elliptical and Fourier-parameterized domains, and on exterior Helmholtz problems over rose curves, show that PPDNO improves accuracy over direct neural operator baselines while adding little inference overhead and generalizing to unseen boundary data. These results suggest that analytic operator structure and geometry-conditioned learning can be combined effectively for boundary solution maps.

In this paper, we devise a $\operatorname{prox}$-based semi-smooth Newton method that is applicable to a finite element discretization of a broad class of nonsmooth convex variational problems, including the TV-minimization problem, the $p$-Dirichlet problem, the obstacle problem, and the elasto-plastic torsion problem. To this end, on the basis of the proximity operator, the discrete primal-dual optimality conditions are reformulated as nonlinear operator equations with Newton-differentiable structure. Under suitable assumptions on the energy densities, we establish the global well-posedness and local super-linear convergence of the resulting semi-smooth Newton method. The proposed approach coincides with established semi-smooth Newton methods for obstacle-type problems, satisfies a primal-dual invariance, and, under suitable additional assumptions, is globally well-posed in the infinite-dimensional setting.