arxiv: 2604.06575 · v2 · submitted 2026-04-08 · 💻 cs.MS · cs.SC· math.AG· math.DG· math.OC

Recognition: no theorem link

Polylab: A MATLAB Toolbox for Multivariate Polynomial Modeling

Shing-Tung Yau, Yi-Shuai Niu

Authors on Pith no claims yet

Pith reviewed 2026-05-10 18:35 UTC · model grok-4.3

classification 💻 cs.MS cs.SCmath.AGmath.DGmath.OC

keywords MATLAB toolboxmultivariate polynomialsGPU computingsymbolic-numeric interfaceautomatic differentiationpolynomial matricesaffine-normal computationperformance benchmarks

0 comments

The pith

Polylab provides a MATLAB toolbox for multivariate polynomials with unified CPU and GPU backends and workload-specific performance guidance.

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

The paper presents Polylab as a toolbox for multivariate polynomial scalars and matrices in MATLAB. It offers a symbolic-numeric interface through three aligned classes for CPU and GPU execution. The software supports construction, algebraic manipulation, simplification, matrix operations, differentiation, Jacobian and Hessian construction, LaTeX export, and interoperability with YALMIP and SOSTOOLS. Recent versions add explicit variable identity handling and affine-normal direction computation via automatic differentiation and log-determinant methods. Benchmarks establish that the standard MPOLY class suits lightweight interactive work while the high-performance GPU class excels at reduction-heavy simplification and medium-to-large affine-normal tasks, with stochastic variants preferred for larger sparse approximations.

Core claim

What carries the argument

The three aligned backend classes MPOLY, MPOLY_GPU and MPOLY_HP that deliver polynomial operations and automatic differentiation for affine-normal direction computation together with MF-logDet-Exact and MF-logDet-Stochastic variants.

If this is right

MPOLY serves as the default for lightweight interactive workloads.
MPOLY-HP becomes advantageous for reduction-heavy simplification.
MPOLY-HP is preferable for medium-to-large affine-normal computation.
The stochastic log-determinant variant is attractive in larger sparse regimes under approximation-oriented parameter choices.
The toolbox enables direct use in optimization workflows through YALMIP and SOSTOOLS interoperability.

Where Pith is reading between the lines

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

Researchers handling mixed polynomial sets can reduce expression errors by adopting the explicit variable naming feature.
The backend selection guidance may allow users to scale polynomial modeling to higher dimensions previously limited by CPU time.
Integration with existing optimization packages suggests the toolbox can accelerate prototyping in control or algebraic-optimization applications.
Future hardware backends could be added while preserving the same user interface.

Load-bearing premise

The reported benchmarks are representative of typical user workloads and that the described features function correctly and reproducibly across different MATLAB versions and hardware configurations.

What would settle it

A user reproducing the benchmark sessions on a different MATLAB release or hardware platform and obtaining performance ratios that reverse the claimed advantages of MPOLY-HP for simplification or affine-normal tasks.

Figures

Figures reproduced from arXiv: 2604.06575 by Shing-Tung Yau, Yi-Shuai Niu.

**Figure 1.** Figure 1: Expanded lightweight core symbolic workloads across MPOLY, MPOLY-GPU, and MPOLY-HP. The workloads use 5-variable [PITH_FULL_IMAGE:figures/full_fig_p015_1.png] view at source ↗

**Figure 2.** Figure 2: Backend crossover on a reduction-heavy simplify workload. The test measures [PITH_FULL_IMAGE:figures/full_fig_p017_2.png] view at source ↗

**Figure 3.** Figure 3: MF-logDet-Exact affine-normal timing across dimensions for MPOLY and MPOLY-HP on sparse quartics with support size 3 [PITH_FULL_IMAGE:figures/full_fig_p018_3.png] view at source ↗

**Figure 4.** Figure 4: Speedup of MF-logDet-Stochastic over MF-logDet-Exact on sparse quartics with support size 2 and [PITH_FULL_IMAGE:figures/full_fig_p019_4.png] view at source ↗

read the original abstract

Polylab is a MATLAB toolbox for multivariate polynomial scalars and polynomial matrices with a unified symbolic-numeric interface across CPU and GPU-oriented backends. The software exposes three aligned classes: MPOLY for CPU execution, MPOLY_GPU as a legacy GPU baseline, and MPOLY_HP as an improved GPU-oriented implementation. Across these backends, Polylab supports polynomial construction, algebraic manipulation, simplification, matrix operations, differentiation, Jacobian and Hessian construction, LaTeX export, CPU-side LaTeX reconstruction, backend conversion, and interoperability with YALMIP and SOSTOOLS. Versions 3.0 and 3.1 add two practically important extensions: explicit variable identity and naming for safe mixed-variable expression handling, and affine-normal direction computation via automatic differentiation, MF-logDet-Exact, and MF-logDet-Stochastic. The toolbox has already been used successfully in prior research applications, and Polylab Version 3.1 adds a new geometry-oriented computational layer on top of a mature polynomial modeling core. This article documents the architecture and user-facing interface of the software, organizes its functionality by workflow, presents representative MATLAB sessions with actual outputs, and reports reproducible benchmarks. The results show that MPOLY is the right default for lightweight interactive workloads, whereas MPOLY-HP becomes advantageous for reduction-heavy simplification and medium-to-large affine-normal computation; the stochastic log-determinant variant becomes attractive in larger sparse regimes under approximation-oriented parameter choices.

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

Polylab documents practical extensions to a MATLAB polynomial toolbox with variable identity handling and affine-normal methods, plus benchmarks that guide backend selection, but it is mainly software documentation.

read the letter

Polylab is a documentation paper for a MATLAB toolbox that adds two practical features in versions 3.0 and 3.1: explicit variable identity and naming to handle mixed variables safely, and affine-normal direction computation using automatic differentiation plus exact and stochastic MF-logDet methods. The paper does well by laying out the architecture clearly, organizing features by typical workflows, and including real MATLAB session examples with outputs. It also reports reproducible benchmarks that compare the CPU-based MPOLY, the legacy GPU, and the improved MPOLY-HP backends. From those, it concludes that MPOLY works best for light interactive tasks, MPOLY-HP shines in simplification-heavy or medium-sized affine-normal calculations, and the stochastic log-det approach helps with bigger sparse cases when approximations are acceptable. Additional strengths include LaTeX export options, derivative computations, and compatibility with YALMIP and SOSTOOLS, plus its track record in actual research use. Soft spots are small and expected for this type of paper. The performance claims rest on the benchmarks, which the authors present as representative, but without more context on how the test cases were chosen, it's possible they don't cover every common workload. There's no new mathematical theory here, just implementation and testing of extensions to an existing core. The descriptions seem consistent internally, with no obvious gaps in the reported functionality. This kind of paper serves MATLAB users in fields like polynomial optimization, algebraic geometry, or control theory who need a unified symbolic-numeric tool with GPU options. A reader looking to adopt or compare polynomial modeling software would get concrete guidance from the examples and recommendations. It deserves a serious referee because software contributions like this can be valuable for the community when documented properly, and the new features address practical pain points. I would recommend sending it to peer review, with perhaps a note to ensure code and benchmark scripts are available for verification.

Referee Report

0 major / 2 minor

Summary. The paper presents Polylab, a MATLAB toolbox for multivariate polynomial scalars and matrices offering a unified symbolic-numeric interface across CPU and GPU backends (MPOLY, MPOLY_GPU, MPOLY_HP). It documents the architecture, user-facing interfaces, supported operations (construction, algebraic manipulation, simplification, matrix operations, differentiation, Jacobian/Hessian construction, LaTeX export), interoperability with YALMIP and SOSTOOLS, and new features in v3.0/3.1 for explicit variable identity/naming and affine-normal direction computation via automatic differentiation, MF-logDet-Exact, and MF-logDet-Stochastic. The manuscript includes example MATLAB sessions with outputs and reports reproducible benchmarks, leading to recommendations on backend selection: MPOLY for lightweight interactive workloads, MPOLY-HP for reduction-heavy simplification and medium-to-large affine-normal tasks, and the stochastic log-determinant variant for larger sparse regimes under approximation-oriented choices.

Significance. If the benchmarks hold, the work delivers a practical, documented toolbox that has already seen research use and provides actionable guidance on performance trade-offs for polynomial modeling tasks common in optimization and control. Strengths include the emphasis on reproducible examples, workflow organization, and extensions for safe mixed-variable handling plus geometry-oriented layers. The paper functions effectively as software documentation with embedded performance data rather than a theoretical contribution.

minor comments (2)

[Benchmarks] The benchmark section would benefit from an explicit statement of the hardware platform, MATLAB version, and problem-size ranges used for the reported timings to help readers judge representativeness for their own workloads.
[Affine-normal computation] In the description of the stochastic log-determinant variant, a short note on the default or recommended approximation parameters (e.g., sample count or tolerance) would improve immediate usability for readers following the example sessions.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for the positive review, the accurate summary of Polylab's features and benchmarks, and the recommendation to accept the manuscript. The referee's description correctly reflects the toolbox architecture, the v3.0/3.1 extensions for explicit variable handling and affine-normal direction computation, and the practical guidance on backend selection.

Circularity Check

0 steps flagged

No significant circularity; paper is software documentation without mathematical derivations

full rationale

This manuscript is a documentation paper for the Polylab MATLAB toolbox. It describes architecture, user interfaces, example sessions, and reports reproducible benchmarks comparing MPOLY, MPOLY-HP, and stochastic log-det backends. No derivation chain, first-principles predictions, or fitted parameters exist that could reduce to inputs by construction. The central performance recommendations follow directly from the presented benchmarks without self-definitional steps, self-citation load-bearing arguments, or ansatz smuggling. The paper is self-contained as software documentation.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

This is a software toolbox documentation paper with no mathematical derivations, physical models, or fitted parameters underlying a central scientific claim.

pith-pipeline@v0.9.0 · 5572 in / 1108 out tokens · 72235 ms · 2026-05-10T18:35:07.449841+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

9 extracted references · 5 canonical work pages · 1 internal anchor

[1]

Lasserre, and Johan Löfberg

Didier Henrion, Jean B. Lasserre, and Johan Löfberg. 2009. GloptiPoly 3: Moments, Optimization and Semidefinite Programming.Optimization Methods and Software24, 4–5 (2009), 761–779

2009
[2]

Hutchinson

Michael F. Hutchinson. 1989. A Stochastic Estimator of the Trace of the Influence Matrix for Laplacian Smoothing Splines.Communications in Statistics - Simulation and Computation18, 3 (1989), 1059–1076

1989
[3]

Johan Löfberg. 2004. YALMIP: A Toolbox for Modeling and Optimization in MATLAB. InProceedings of the IEEE International Symposium on Computer Aided Control Systems Design. IEEE, Taipei, Taiwan, 284–289. doi:10.1109/CACSD.2004.1393890

work page doi:10.1109/cacsd.2004.1393890 2004
[4]

Yi-Shuai Niu and Roland Glowinski. 2022. Discrete Dynamical System Approaches for Boolean Polynomial Optimization.Journal of Scientific Computing92, 2 (2022), 46

2022
[5]

Yi-Shuai Niu, Artan Sheshmani, and Shing-Tung Yau. 2026. Affine Normal Directions via Log-Determinant Geometry: Scalable Computation under Sparse Polynomial Structure. https://arxiv.org/abs/2604.01163 arXiv:2604.01163

work page arXiv 2026
[6]

Yi-Shuai Niu, Artan Sheshmani, and Shing-Tung Yau. 2026. Yau’s Affine Normal Descent: Algorithmic Framework and Convergence Analysis. https://arxiv.org/abs/2603.28448 arXiv:2603.28448

work page internal anchor Pith review Pith/arXiv arXiv 2026
[7]

Yi-Shuai Niu, Hoai An Le Thi, and Dinh Tao Pham. 2024. On Difference-of-SOS and Difference-of-Convex-SOS Decompositions for Polynomials. SIAM Journal on Optimization34, 2 (2024), 1852–1878

2024
[8]

Yi-Shuai Niu, Yajuan Wang, Hoai An Le Thi, and Dinh Tao Pham. 2019. High-order Moment Portfolio Optimization via An Accelerated Difference- of-Convex Programming Approach and Sums-of-Squares. https://arxiv.org/abs/1906.01509 arXiv:1906.01509

work page arXiv 2019
[9]

Antonis Papachristodoulou, James Anderson, Gavin Valmorbida, Stephen Prajna, Petter Seiler, and Pablo A. Parrilo. 2013. SOSTOOLS: Sum of Squares Optimization Toolbox for MATLAB. https://arxiv.org/abs/1310.4716 arXiv:1310.4716

work page arXiv 2013