betaFeedback

WorksMathematics

Porteous Meets Walsh: Counting Critical Points for H2\mathcal{H}_2-Norm Model Order Reduction Using Algebraic Geometry Behind Multiparameter Eigenvalue Problems

Abstract

Several polynomial systems arising in systems and control have been shown to exhibit a partially linear structure, which allows them to be reformulated as a multiparameter eigenvalue problem (MEP). In this work, we show how such MEP formulations can be leveraged to obtain refined information about the solution set of the polynomial system. We demonstrate, in the setting of single-input single-output H2\mathcal{H}_2-norm model order reduction, how the Thom–Porteous formula can be used to derive an exact count of the complex solutions of the critical equations derived from Walsh's theorem. We prove, under suitable genericity assumptions, that their set of complex solutions is finite and that all solutions occur without multiplicity. Moreover, we derive an explicit formula for the total number of complex solutions as a function of the model orders. Throughout, we provide a self-contained and application-oriented exposition of the algebro-geometric tools required for our analysis, with the aim of making them accessible to the broader engineering community.

Introduction

Systems of polynomial equations arise frequently in engineering applications; still, solving them comprehensively—numerically or symbolically—is computationally challenging. For such systems, one typically has access to unconditional but only coarse upper bounds on the number of affine solutions, such as those provided by Bézout’s theorem. Nevertheless, having a precise a priori count of the solutions is valuable both for the design and use of computational methods. It provides a concrete proxy for the expected computational cost and can be used to validate computations by confirming that the full solution set has been recovered.

A multiparameter eigenvalue problem (MEP) generalizes the standard eigenvalue problem (A−λI)v=0(A-\lambda I)v=0 by allowing multiple spectral parameters and by dropping the restriction of linearity in the parameters [2, 47, 11]. For rectangular MEPs [45, 23], one looks for all λ=(λ1,…,λk)∈Ck\lambda=(\lambda_1,\ldots,\lambda_k)\in\mathbb{C}^k and nonzero v∈Cmv\in\mathbb{C}^m such that

M(λ)v=0,(1)M(\lambda)v=0, \tag*{(1)}

where M(λ)M(\lambda) is a tall, ℓ×m\ell\times m matrix whose entries are polynomials in λ\lambda. The solutions λ\lambda are called eigentuples [11] and vv the corresponding eigenvector. For k=1k=1 and MM linear in λ\lambda, we recover the (generalized) eigenvalue problem.

Several systems of polynomial equations arising in applications from systems and control have been shown to exhibit a partially linear structure, such that they can be reformulated as a rectangular MEP, see for example [6, 44, 13, 3, 33]. In these cases, instead of solving the system directly using general-purpose polynomial system solvers (e.g., [9]), one can alternatively obtain the solutions by computing the eigentuples of the corresponding MEP using specialized solvers (e.g., [45, 23, 46]).

In this paper, we argue that the reformulation of the system of polynomial equations as an MEP can also yield more theoretical structural insights into the problem at hand. In particular, two central results in algebraic geometry, when applied in the context of these MEP reformulations, can be used to obtain finer information about the solution set of the polynomial system. When a polynomial system is expected to have finitely many solutions, the fiber dimension theorem is a useful tool to verify this. Consequently, where finiteness of the solutions can be established, the Thom–Porteous formula, applied to the determinantal variety associated with the MEP reformulation, can often be used to derive an exact count of the affine complex solutions.

To ground our discussion, we illustrate the use of these tools in the context of a problem from systems and control: the H2\mathcal{H}_2-norm model reduction problem for single-input single-output (SISO) linear time-invariant (LTI) models, in which a given model is approximated by one of lower order. It corresponds to a non-convex optimization problem and may therefore admit multiple local optima, which can be characterized as (a subset of the) real-valued common roots of a system of polynomial equations [15, 39, 24, 31]. These equations, commonly referred to as the critical equations, follow from Walsh’s Theorem [48], which, together with the Thom–Porteous formula, motivates the title of this paper.

We study the complex solution set of the critical equations as a whole and refer to its elements as the critical points. When the given higher-order model is sufficiently generic, the set of critical points is known to be finite [Bar86, Theorem 3]; in fact, simple poles suffice for this to hold [HPB12, Theorem 3.2]. This observation has motivated several approaches that use algebraic and numerical linear algebra methods to enumerate all such critical points; see, e.g., [24, 37, 1, 3, 31]. When complete, such enumerations allow one to certify global optimality. However, they are computationally feasible only for relatively small problems and are therefore best regarded as theoretical tools, mostly used for benchmarking purposes. Practical solution approaches, such as the iterative rational Krylov algorithm (IRKA), scale to large problems by iteratively enforcing the first-order optimality conditions, but converge at best to a single local optimum, without any guarantee of global optimality; see, e.g., [19] and references therein. Despite the extensive literature on H2\mathcal{H}_2-norm model reduction, an exact count of the critical points for arbitrary problem sizes is, to the best of our knowledge, not available. For the co-order one case, i.e., when the order is reduced by one, [24] provides an upper bound.

Recent work showed that the critical equations allow for an MEP reformulation [31]. By applying the algebro-geometric tools outlined above, we arrive at the following result:

Theorem (Theorem 3.10). For generic higher-order models of McMillan degree NN, the number of critical points of the model order reduction problem to degree nn, 1≤n<N1 \le n < N, is

∑j=0n(N−n−1+jj)22j.\sum_{j=0}^{n} \binom{N-n-1+j}{j} 2^{2j}.

The main contributions of our present work are as follows:

  1. We give a new proof that, for higher-order models with simple poles, the set of critical points is finite (Corollary 3.9). Moreover, we show for generic higher-order models that the solutions of the critical equations occur without multiplicity, a property favorable for numerical computation (Corollary 3.5).

  1. Assuming the higher-order model has only simple poles, we derive an explicit formula for the number of critical points, that is, the total number of complex solutions to the critical equations, as a function of the full and reduced orders (Theorem 3.10). For higher-order models with one or more multiple poles, we illustrate how the aforementioned count can be

used to derive an upper bound on the number of critical points (§ 4.1). Although we consider the continuous-time case throughout, our analysis easily extends to a discrete-time setting (§ 4.2).

(3) We provide a self-contained, application-oriented exposition of the key algebro-geometric tools, making them accessible to a broader engineering audience (§ 3 and Appendix § 8). These techniques are not specific to the model reduction problem, but are equally applicable to other engineering problems where such MEP formulations arise.

We would like to note that our results connect to a broader line of work in algebraic geometry. Counting of critical points is related to the Euclidean distance (ED) degree of an algebraic variety, defined as the number of complex critical points of the squared Euclidean distance from a generic data point to the smooth part of the variety [14]. In discrete time, the squared H2\mathcal{H}_2-norm of a SISO LTI system is the sum of its squared impulse response coefficients. The discrete-time model reduction problem can be seen as an infinite-dimensional instance of structured low-rank approximation of Hankel matrices, whose ED degrees are studied in [35 SP25] and [29]. The latter employs similar algebro-geometric tools as used here, and can be seen as a companion to the present paper.

Outline The remainder of the paper is organized as follows. In § 2, we give a formal statement of the model reduction problem and derive the critical equations of the optimization problem. The heart of the paper is § 3, where we lay out a recipe for establishing finiteness and deriving a count of the solution set of MEPs. We implement this recipe for the model reduction problem and derive the corresponding solution count. Several further considerations are given in § 4: the special case where the given higher-order model has multiple poles and the discrete-time setting. We illustrate all theoretical results by means of several examples in § 5 and conclude the paper in § 6. The complementary Appendix § 8 gives an elementary introduction to the essential algebraic geometry concepts relevant to our work.

Notation Throughout, scalars are denoted by lowercase letters (e.g., s∈Cs \in\mathbb{C}), vectors by boldface lowercase letters (e.g., a∈Ck\mathbf{a} \in\mathbb{C}^k), and matrices by boldface uppercase letters (e.g., M∈Ck×k\mathbf{M} \in\mathbb{C}^{k \times k}). Let C[s]≤k\mathbb{C}[s]_{\leq k} denote the (k+1)(k+1)-dimensional vector space of complex polynomials in ss of degree at most kk. For a polynomial f=∑i=0kfisi∈C[s]≤kf=\sum_{i=0}^{k}f_i s^i\in\mathbb{C}[s]_{\leq k}, we write its coefficient vector in boldface as f=(f0,…,fk)T∈Ck+1\mathbf{f}=(f_0,\ldots,f_k)^{\mathsf{T}}\in\mathbb{C}^{k+1} and, conversely, any vector f∈Ck+1\mathbf{f}\in\mathbb{C}^{k+1} determines a polynomial f∈C[s]≤kf\in\mathbb{C}[s]_{\leq k}. Let Tℓ(f)∈Cℓ×(ℓ+k)T_{\ell}(f)\in\mathbb{C}^{\ell\times(\ell+k)} denote the banded Toeplitz matrix generated by ℓ\ell forward-shifts of the coefficient vector f∈Ck+1\mathbf{f}\in\mathbb{C}^{k+1}:

Tℓ(f):=[f0⋯fk−1fk0⋯00f0⋯fk−1fk⋯0⋮⋱⋱⋱⋱⋱⋮0⋯0f0⋯fk−1fk].T_{\ell}(f):= \begin{bmatrix} f_0 & \cdots& f_{k-1} & f_k & 0 & \cdots& 0\\ 0 & f_0 & \cdots& f_{k-1} & f_k & \cdots& 0\\ \vdots& \ddots& \ddots& \ddots& \ddots& \ddots& \vdots\\ 0 & \cdots& 0 & f_0 & \cdots& f_{k-1} & f_k \end{bmatrix}.

Toeplitz matrices constructed in this way allow us to represent polynomial multiplication in terms of coefficient vectors. For instance, if h∈C[s]h\in\mathbb{C}[s], then the product hfhf has coefficient vector Tk+1(h)TfT_{k+1}(h)^{\mathsf{T}}\mathbf{f}. For f∈C[s]≤kf\in\mathbb{C}[s]_{\leq k}, write fh:=tkf(s/t)∈C[s,t]kf^{\mathsf{h}}:=t^k f(s/t)\in\mathbb{C}[s,t]_k for its homogenization to degree kk. Algebraic varieties are denoted by uppercase letters, with calligraphic font occasionally used for emphasis (e.g., W\mathcal{W}). Unless stated otherwise, all open and closed subsets mentioned in this paper are taken with respect to the Zariski topology.

H₂-norm model order reduction

Throughout, we work with transfer functions of real, continuous-time, strictly causal, stable, single-input single-output (SISO) linear time-invariant (LTI) systems of finite order (but see Section 4 for the discrete-time setting). For a fixed integer N≥1N \ge1, let MN\mathcal{M}_N denote the set of all such transfer functions of McMillan degree exactly NN. Any model H∈MNH \in\mathcal{M}_N can be parametrized as

H(s)=dc=dN−1sN−1+⋯+d1s+d0cNsN+cN−1sN−1+⋯+c1s+c0.H(s) = \frac{d}{c} = \frac{d_{N-1}s^{N-1} + \cdots+ d_1s + d_0}{c_Ns^N + c_{N-1}s^{N-1} + \cdots+ c_1s + c_0}.

where c∈R[s]≤Nc \in\mathbb{R}[s]_{\le N} and d∈R[s]≤N−1d \in\mathbb{R}[s]_{\le N-1} are real polynomials with cN≠0c_N \ne0 and gcd⁡(c,d)=1\operatorname{gcd}(c,d) = 1. (Hurwitz-)Stability here means that all poles of HH, the roots of cc, lie in the open left half-plane. Strict causality means deg⁡c>deg⁡d\deg c > \deg d, so that the impulse response vanishes for negative time and is square-integrable for t≥0t \ge0. Consequently, elements of MN\mathcal{M}_N are real rational functions in the Hardy space H2\mathcal{H}_2 of functions analytic in the open right half-plane and square-integrable along the imaginary axis.

Given a target reduced order 1≤n<N1 \le n < N, we seek an approximant H^∈Mn\widehat{H} \in\mathcal{M}_n, parametrized as

H^(s)=ba=bn−1sn−1+⋯+b1s+b0ansn+an−1sn−1+⋯+a1s+a0,\widehat{H}(s) = \frac{b}{a} = \frac{b_{n-1}s^{n-1} + \cdots+ b_1s + b_0}{a_ns^n + a_{n-1}s^{n-1} + \cdots+ a_1s + a_0},

where a∈R[s]≤na \in\mathbb{R}[s]_{\le n} and b∈R[s]≤n−1b \in\mathbb{R}[s]_{\le n-1} are real polynomials with an≠0a_n \ne0, gcd⁡(a,b)=1\operatorname{gcd}(a,b) = 1, and aa Hurwitz stable. The goal of the model order reduction problem under consideration is to minimize the H2\mathcal{H}_2-norm of the approximation error H−H^H-\widehat{H}. Formally, the problem corresponds to

minimize⁡H^∈Mn J2,J2=∥H−H^∥H22=12π∫−∞∞∣H(jω)−H^(jω)∣2 dω=∫0∞∣h(t)−h^(t)∣2 dt,(2)\begin{aligned} \underset{\widehat{H}\in\mathcal{M}_n}{\operatorname{minimize}}\ J^2,\qquad J^2 = \lVert H-\widehat{H}\rVert_{\mathcal{H}_2}^2 = \frac{1}{2\pi}\int_{-\infty}^{\infty}\lvert H(\mathrm{j}\omega)-\widehat{H}(\mathrm{j}\omega)\rvert^2\,\mathrm{d}\omega\\ &= \int_0^\infty\lvert h(t)-\widehat{h}(t)\rvert^2\,\mathrm{d}t, \tag*{(2)} \end{aligned}

where h(t)h(t) and h^(t)\widehat{h}(t) denote the impulse responses of HH and H^\widehat{H}, respectively, obtained via the inverse Laplace transform.

We restrict the search to reduced models of McMillan degree exactly nn, rather than at most nn. This is because the set Mn\mathcal{M}_n is open and dense in M≤n:=M0∪M1∪⋯∪Mn\mathcal{M}_{\le n}:=\mathcal{M}_0\cup\mathcal{M}_1\cup\cdots\cup\mathcal{M}_n, where M0:={0}\mathcal{M}_0:=\{0\}, so that every model of lower McMillan degree is a limit of models of degree nn. Moreover, every local minimizer of (2), when it were to be considered over M≤n\mathcal{M}_{\le n}, is guaranteed to have McMillan degree exactly nn [39 Theorem 5.3].1 The set Mn\mathcal{M}_n is a smooth manifold of dimension 2n2n: indeed, normalize aa to be monic so that an=1a_n=1 and the remaining coefficients of aa and bb range over an open subset of R2n\mathbb{R}^{2n}, obtained by removing those points where aa and bb share one or more common factors. We call H^∈Mn\widehat{H}\in\mathcal{M}_n a stationary point of (2) if the gradient of J2J^2 with respect to these 2n2n coefficients vanishes.

References

  1. [1]F. F. Alsubaie. “H₂ Optimal Model Reduction for Linear Dynamical Systems and the Solution of Multiparameter Matrix Pencil Problems”. PhD thesis. Imperial College London, 2019 (cit. on p. 2).DOI
  2. [2]F.V. Atkinson. Multiparameter Eigenvalue Problems. Volume I: Matrices and Compact Operators. Vol. 82. Mathematics in Science and Engineering. New York: Academic Press, 1972 (cit. on p. 1).
  3. [3]Oscar Mauricio Agudelo, Christof Vermeersch, and Bart De Moor. “Globally Optimal H2-Norm Model Reduction: A Numerical Linear Algebra Approach”. In: IFAC-PapersOnLine 54.9 (2021), pp. 564–571. ISSN: 2405-8963. DOI: 10.1016/j.ifacol.2021.06.117 (cit. on pp. 1, 2, 17).DOI
  4. [4]P. R. Aigrain and E. M. Williams. “Synthesis of n-Reactance Networks for Desired Transient Response”. In: Journal of Applied Physics 20.6 (1949), pp. 597–600 (cit. on p. 4).DOI
  5. [6]Kim Batselier, Philippe Dreesen, and Bart De Moor. “Prediction Error Method Identification is an Eigenvalue Problem”. In: IFAC Proceedings Volumes 45.16 (July 2012), pp. 221–226. ISSN: 1474-6670. DOI: 10.3182/20120711-3-BE-2027.00132 (cit. on p. 1).DOI
  6. [7]I. Bleylevens. “Algebraic Polynomial System Solving and Applications”. PhD thesis. Maastricht, The Netherlands: Maastricht University, 2010. DOI: 10.26481/dis.20101209ib (cit. on p. 5).DOI
  7. [8]L. Baratchart, M. Olivi, and F. Wielonsky. “On a rational approximation problem in the real Hardy space H2”. In: Theoretical Computer Science 94.2 (Mar. 1992), pp. 175–197. ISSN: 0304-3975. DOI: 10.1016/0304-3975(92)90033-C (cit. on pp. 20, 21).DOI
  8. [9]P. Breiding and S. Timme. “HomotopyContinuation.jl: A Package for Homotopy Continuation in Julia”. In: International Congress on Mathematical Software. 2018, pp. 458–465 (cit. on p. 2).arxiv.org/abs/1711.10911
  9. [10]David A. Cox, John Little, and Donal O’Shea. Ideals, Varieties, and Algorithms: An Introduction to Computational Algebraic Geometry and Commutative Algebra. Springer International Publishing, 2015. ISBN: 9783319167213. DOI: 10.1007/978-3-319-16721-3 (cit. on p. 22).
  10. [11]Katrien De Cock and Bart De Moor. “Multiparameter Eigenvalue Problems and Shift-invariance”. In: IFAC-PapersOnLine 54.9 (2021), pp. 159–165. ISSN: 2405-8963. DOI: 10.1016/j.ifacol.2021.06.071 (cit. on p. 1).DOI
  11. [12]Emily Clader and Dustin Ross. Beginning in Algebraic Geometry. Springer Nature Switzerland, 2025. DOI: 10.1007/978-3-031-88819-9 (cit. on pp. 7, 22).DOI
  12. [13]B. De Moor. “Least squares optimal realisation of autonomous LTI systems is an eigenvalue problem”. In: Communications in Information and Systems 20.2 (2020), pp. 163–207. ISSN: 2163-4548. DOI: 10.4310/cis.2020.v20.n2.a4 (cit. on p. 1).DOI
  13. [14]Jan Draisma, Emil Horobet, Giorgio Ottaviani, Bernd Sturmfels, and Rekha R. Thomas. “The Euclidean Distance Degree of an Algebraic Variety”. In: Foundations of Computational Mathematics 16.1 (Jan. 2015), pp. 99–149. ISSN: 1615-3383. DOI: 10.1007/s10208-014-9240-x (cit. on p. 3).DOI
  14. [15]B. De Moor, P. Van Overschee, and G. Schelfhout. “H₂ Model Reduction for SISO Systems”. In: vol. 26. 2. Elsevier BV, July 1993, pp. 895–898. DOI: 10.1016/s1474-6670(17)49262-7 (cit. on p. 2).DOI
  15. [16]David Eisenbud and Joe Harris. 3264 and All That: A Second Course in Algebraic Geometry. Cambridge University Press, Mar. 2016. ISBN: 9781107602724. DOI: 10.1017/cbo9781139062046 (cit. on pp. 7, 22, 27).DOI
  16. [17]David Eisenbud. Commutative Algebra: with a View Toward Algebraic Geometry. Springer New York, 1995. ISBN: 9781461253501. DOI: 10.1007/978-1-4612-5350-1 (cit. on p. 25).DOI
  17. [18]William Fulton. Intersection Theory. Springer New York, 1998. ISBN: 9781461217008. DOI: 10.1007/978-1-4612-1700-8 (cit. on p. 26).DOI
  18. [19]S. Gugercin, A. C. Antoulas, and C. Beattie. “H₂ Model Reduction for Large-Scale Linear Dynamical Systems”. In: SIAM Journal on Matrix Analysis and Applications 30.2 (Jan. 2008), pp. 609–638. ISSN: 1095-7162. DOI: 10.1137/060666123 (cit. on pp. 2, 4).
  19. [20]Giovanni Z. Giambelli. “Sul luogo dei punti di contatto delle ipersuperficie di due dati sistemi lineari”. In: Rendiconti del Circolo Matematico di Palermo 23.1 (Dec. 1907), pp. 302–306. ISSN: 1973-4409. DOI: 10.1007/bf03013525 (cit. on p. 7).DOI
  20. [21]Robin Hartshorne. Algebraic Geometry. Springer New York, 1977. ISBN: 9781475738490. DOI: 10.1007/978-1-4757-3849-0 (cit. on pp. 22, 24).DOI
  21. [22]Joe Harris. Algebraic Geometry: A First Course. Springer New York, 1992. ISBN: 9781475721898. DOI: 10.1007/978-1-4757-2189-8 (cit. on p. 22).DOI
  22. [23]Michiel E. Hochstenbach, Tomaž Košir, and Bor Plestenjak. “Numerical methods for rectangular multiparameter eigenvalue problems, with applications to finding optimal ARMA and LTI models”. In: Numerical Linear Algebra with Applications 31.2 (Nov. 2023). ISSN: 1099-1506. DOI: 10.1002/nla.2540 (cit. on pp. 1, 2, 9).DOI
  23. [24]B. Hanzon, J. Maciejowski, and C. T. Chou. Optimal H2 order-one reduction by solving eigenproblems for polynomial equations. 2007. arXiv: 0706.1862 [math.OC] (cit. on pp. 2, 5, 12).arxiv.org/abs/0706.1862
  24. [26][HYT01] Xue-Xiang Huang, Wei-Yong Yan, and K. L. Teo. “A new approach to frequency weighted L2 optimal model reduction”. In: International Journal of Control 74.12 (Jan. 2001), pp. 1239–1246. ISSN: 1366-5820. DOI: 10.1080/00207170110063499 (cit. on p. 21).DOI
  25. [27][Kay26] E. L. Kayser. “Computation and Complexity in Algebraic Geometry”. PhD thesis. Universität Leipzig, 2026. URL: https://nbn-resolving.org/urn:nbn:de:bsz:15-qucosa2–1067809 (cit. on pp. 16, 21, 28).
  26. [28][KL26] E. L. Kayser and S. Lagauw. Code repository accompanying the paper “Porteous Meets Walsh: Counting Critical Points for H2-Norm Model Reduction Using Algebraic Geometry Behind Multiparameter Eigenvalue Problems”. 2026. URL: https://github.com/leokayser/h2mor (cit. on p. 17).
  27. [29][KL26+] E. L. Kayser and S. Lagauw. “Generic and specific Euclidean distance degrees of secant varieties to the rational normal curve”. Manuscript in preparation. 2026 (cit. on pp. 3, 7, 21).
  28. [30][Lag26] S. Lagauw. “Global Solution to Problems in System Identification and Model Order Reduction”. PhD thesis. KU Leuven, 2026 (cit. on p. 21).
  29. [31][LAM23] Sibren Lagauw, Oscar Mauricio Agudelo, and Bart De Moor. “Globally Optimal SISO H2-Norm Model Reduction Using Walsh’s Theorem”. In: IEEE Control Systems Letters 7 (2023), pp. 1670–1675. ISSN: 2475-1456. DOI: 10.1109/lcsys.2023.3276299 (cit. on pp. 2, 9, 17).DOI
  30. [32][Lax59] Peter D. Lax. “Translation invariant spaces”. In: Acta Mathematica 101.3-4 (1959), pp. 163–178. ISSN: 0001-5962. DOI: 10.1007/bf02559553 (cit. on p. 5).DOI
  31. [33][LVD26] Sibren Lagauw, Lukas Vanpoucke, and Bart De Moor. “Exact solution to the least squares realization problem as a multiparameter eigenvalue problem”. In: Automatica 185 (Mar. 2026), p. 112792. ISSN: 0005-1098. DOI: 10.1016/j.automatica.2025.112792 (cit. on p. 1).DOI
  32. [34][ML67] L. Meier and D. Luenberger. “Approximation of linear constant systems”. In: IEEE Transactions on Automatic Control 12.5 (Oct. 1967), pp. 585–588. ISSN: 1558-2523. DOI: 10.1109/tac.1967.1098680 (cit. on pp. 4, 6).DOI
  33. [35][OSS14] Giorgio Ottaviani, Pierre-Jean Spaenlehauer, and Bernd Sturmfels. “Exact Solutions in Structured Low-Rank Approximation”. In: SIAM Journal on Matrix Analysis and Applications 35.4 (Jan. 2014), pp. 1521–1542. ISSN: 1095-7162. DOI: 10.1137/13094520x (cit. on p. 3).DOI
  34. [36][Par04] J. Partington. Linear Operators and Linear Systems: An Analytical Approach to Control Theory. Vol. 60. London Mathematical Society Student Texts. Cambridge University Press, 2004. DOI: 10.1017/CBO9780511616693 (cit. on pp. 5, 16).DOI
  35. [37][PBH12] Ralf Peeters, Ivo Bleylevens, and Bernard Hanzon. “On algebraic and linear algebraic aspects of co-order three H2-model order reduction”. In: IFAC Proceedings Volumes 45.16 (2012), pp. 704–709. ISSN: 1474-6670. DOI: 10.3182/20120711–3–be–2027.00379 (cit. on p. 2).DOI
  36. [38][Por71] I. R. Porteous. “Simple singularities of maps”. In: Proceedings of Liverpool Singularities — Symposium I. Springer Berlin Heidelberg, 1971, pp. 286–307. ISBN: 9783540365310. DOI: 10.1007/bfb0066829 (cit. on p. 7).DOI
  37. [39]Phillip A. Regalia. Adaptive IIR Filtering in Signal Processing and Control. Routledge, Apr. 2018. ISBN: 9781315136653. DOI: 10.1201/9781315136653 (cit. on pp. 2, 4, 5, 20, 21).DOI
  38. [40]Igor R. Shafarevich. Basic Algebraic Geometry 1: Varieties in Projective Space. Springer Berlin Heidelberg, 2013. ISBN: 9783642379567. DOI: 10.1007/978-3-642-37956-7 (cit. on pp. 7, 22, 24, 26).DOI
  39. [42]Andrew J. Sommese and Charles W. Wampler. The Numerical Solution of Systems of Polynomials Arising in Engineering and Science. World Scientific, Mar. 2005. ISBN: 9789812567727. DOI: 10.1142/5763 (cit. on p. 22).DOI
  40. [43]Paul Van Dooren, Kyle A. Gallivan, and P.-A. Absil. “H₂-Optimal Model Reduction with Higher-Order Poles”. In: SIAM Journal on Matrix Analysis and Applications 31.5 (Jan. 2010), pp. 2738–2753. ISSN: 1095-7162. DOI: 10.1137/080731591 (cit. on pp. 5, 20).DOI
  41. [44]Christof Vermeersch and Bart De Moor. “Globally Optimal Least-Squares ARMA Model Identification is an Eigenvalue Problem”. In: IEEE Control Systems Letters 3.4 (Oct. 2019), pp. 1062–1067. ISSN: 2475-1456. DOI: 10.1109/lcsys.2019.2920765 (cit. on p. 1).DOI
  42. [45]Christof Vermeersch and Bart De Moor. “Two Double Recursive Block Macaulay Matrix Algorithms to Solve Multiparameter Eigenvalue Problems”. In: IEEE Control Systems Letters 7 (2023), pp. 319–324. ISSN: 2475-1456. DOI: 10.1109/lcsys.2022.3182632 (cit. on pp. 1, 2, 9).DOI
  43. [46]C. Vermeersch and B. De Moor. Solving Multivariate Polynomial Systems and Rectangular Multiparameter Eigenvalue Problems with MacaulayLab. Tech. rep. Leuven, Belgium: KU Leuven, 2025 (cit. on pp. 2, 9).arxiv.org/abs/2605.20884
  44. [47]Hans Volkmer. Multiparameter Eigenvalue Problems and Expansion Theorems. Springer Berlin Heidelberg, 1988. ISBN: 9783540460152. DOI: 10.1007/bfb0089295 (cit. on p. 1).DOI
  45. [48]J. L. Walsh. Interpolation and Approximation by Rational Functions in the Complex Domain. 3rd. American Mathematical Society, 1960 (cit. on pp. 2, 5).

Paper details

vlnc:W_ueh9a4vz
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 10

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

GenAI-based tools

View statement
Looking for related papers…
Contents