betaFeedback

WorksMathematics

Variational Splines and Generalized Gaussian Processes

Abstract

Tikhonov regularization, kriging, and Gaussian-process regression share the same algorithm because, for a linear inverse problem with a Gaussian prior and Gaussian noise, the minimum-mean-square-error (MMSE) estimator is the solution of a regularized least-squares problem. We show that this equivalence survives the passage from vectors to generalized functions: the regularized least-squares estimators become generalized splines, and the Gaussian random vectors become generalized Gaussian processes (GGPs) on the dual of a nuclear space. A whitening operator L:S→L2\mathrm{L}: \mathcal{S} \to L_2 specifies both the native Hilbert space HL⊂S′\mathcal{H}_\mathrm{L} \subset \mathcal{S}' of the splines and a family of GGPs, which solve a stochastic differential equation subject to boundary conditions. We prove that the Gaussian spaces of these processes are congruent to the core L∗(L2)\mathrm{L}^\ast(L_2) of the native space, and that their MMSE estimators from finitely many linear measurements are generalized splines. Our framework extends the classical correspondences of Kimeldorf-Wahba, Parzen and Kailath in two respects: our estimators form a broader class that includes generalized functions, and we give a proper Bayesian treatment of the case where the null space of L\mathrm{L} is non-trivial. The framework gives a precise meaning to statements such as ``fractional splines are the optimal estimators of fractional Brownian motion,'' and it provides a rigorous Bayesian formulation of infinite-dimensional linear inverse problems.

Introduction

The equivalence between optimal estimation under the Gaussian hypothesis and regularized least squares is one of the organizing principles of estimation theory. It is also the reason why the same algorithm circulates under different names in different communities: Tikhonov regularization [55] and ridge regression [18] in inverse problems and statistics, the Wiener filter in signal processing [24, 67], kriging in geostatistics [35, 39], and Gaussian-process regression in machine learning [5, 44]. In finite dimensions, the equivalence is elementary. Let us consider the linear measurement model y=Hx+ϵ\mathbf{y} = \mathbf{H}\mathbf{x} + \boldsymbol{\epsilon}, where the signal x∈RN\mathbf{x} \in\mathbb{R}^{N} is a realization of a Gaussian random vector X∼N(0,CX)\mathbf{X} \sim\mathcal{N}(\mathbf{0}, \mathbf{C}_{X}) and where ϵ∼N(0,σ2I)\boldsymbol{\epsilon} \sim\mathcal{N}(\mathbf{0}, \sigma^{2}\mathbf{I}) is an independent measurement noise. The posterior distribution of X\mathbf{X} given y\mathbf{y} is Gaussian, so that its mean and its mode coincide: the minimum-mean-square-error (MMSE) estimator and the maximum-a-posteriori (MAP) estimator are one and the same. The MAP estimator minimizes the negative log-posterior, which makes it the solution of a penalized least-squares problem. What is specific to the Gaussian hypothesis is that the penalty is quadratic and that the minimizer is also the MMSE estimator: the common estimator is the solution of the Tikhonov problem

xλ=arg min⁡x∈RN(∥y−Hx∥22+λ∥Lx∥22)(1)\mathbf{x}_{\lambda} = \underset{\mathbf{x}\in\mathbb{R}^{N}}{\operatorname{arg\,min}}\left(\lVert\mathbf{y}-\mathbf{H}\mathbf{x}\rVert_{2}^{2}+\lambda\lVert\mathbf{L}\mathbf{x}\rVert_{2}^{2}\right) \tag*{(1)}

with λ=σ2\lambda= \sigma^{2}, provided that the regularization matrix L\mathbf{L} whitens the prior, in the sense that LCXLT=I\mathbf{L}\mathbf{C}_{X}\mathbf{L}^{T} = \mathbf{I}. This classic correspondence is the template that the present paper transposes to the continuum. Our deterministic reference is the formulation of the variational theory of splines in [59, 60], which is the most extensive that we are aware of: the regularization operator L\mathbf{L} is any spline-admissible operator, whose null space may be nontrivial, the signals are generalized functions, and the measurements are linear functionals that are continuous on the native space. Our aim is to establish the correspondence at this level of generality. The generalized splines take the role of the linear estimators: they are the solutions of the continuous-domain counterparts of (1), and they are of the form f=∑m=1MamA{νm}+pf = \sum_{m=1}^{M} a_{m}A\{\nu_{m}\} + p, where the νm\nu_{m} are the measurement functionals, AA is the Green’s operator (conditional inverse) of the normal operator L∗L\mathbf{L}^{*}\mathbf{L}, and pp lies in the null space of L\mathbf{L}. The generalized Gaussian processes (GGPs) take the role of the Gaussian random vectors; they are random elements of the dual S′\mathcal{S}^{\prime} of a nuclear Gelfand pair (S,S′)(\mathcal{S},\mathcal{S}^{\prime}), in the sense of Gelfand and Vilenkin [15]; the prototype of such a pair is the Schwartz space of rapidly decaying test functions together with its dual, the space of tempered distributions.

The continuous-domain version of this correspondence has a long history. It was uncovered by Kimeldorf and Wahba, first for the case where L\mathbf{L} is shift-invariant with a trivial null space [26] and then for the Tchebycheffian splines on an interval [27], and it was developed into a complete statistical methodology in [50, 63]. In parallel, Parzen [41] recast time-series estimation in the language of reproducing-kernel Hilbert spaces (RKHS) [1], while Kailath and his students connected RKHS norms, innovations representations, and Fredholm equations [22, 23, 64, 65]. The modern kernel methods of machine learning inherit the same duality [19, 25, 40, 49], as does the stochastic-partial-differential-equation approach to spatial statistics [30, 31], in which a Gaussian field is specified through the operator that whitens it rather than through its covariance.

There is, however, a loose end in this otherwise satisfying picture, and it affects precisely the estimators that approximation theory favors. When the regularization operator LL has a nontrivial null space Np=span⁡{p1,…,pN0}\mathcal{N}_p = \operatorname{span}\{p_1,\ldots,p_{N_0}\}—which is the case of the derivative operators DmD^m that generate the polynomial splines, the workhorses of the field—the LL-smoothing spline loses its MMSE optimality. What survives is the weaker statement that it is the best linear unbiased estimator in a so-called fixed-effect model, where the bias is measured relative to a fixed but unknown drift p∈Npp \in\mathcal{N}_p [63 Sec. 1.5]. Geostatistics meets the same obstacle: the intrinsic random functions of order kk of Matheron [35] are defined only modulo the polynomials of degree kk, and universal kriging is, likewise, a best linear unbiased predictor. This is frustrating, for these are the estimators with the best deterministic approximation rates, and they owe those rates to the very null space that degrades their statistical status. The reason for the degradation is easy to identify: the process whose covariance operator is the inverse of L∗LL^*L does not exist when LL is not invertible, so that there is no prior with respect to which a Bayesian estimator could be optimal.

The resolution that we propose is to construct the missing prior, in a setting that is broad enough to accommodate signals that are generalized functions. The conditioned parametrization of [60] provides the two ingredients: a right-inverse Lϕ−1L_\phi^{-1} of LL that is made unique by N0N_0 boundary conditions ϕ(f)=(⟨ϕ1,f⟩,…,⟨ϕN0,f⟩)=0\phi(f) = (\langle\phi_1,f\rangle,\ldots,\langle\phi_{N_0},f\rangle) = \mathbf{0}, and the associated Gram operator Aϕ=Lϕ−1Lϕ−∗A_\phi= L_\phi^{-1}L_\phi^{-*}, which is symmetric and pp-conditionally positive-definite: its quadratic form is positive on the test functions that annihilate Np\mathcal{N}_p, while it vanishes on the boundary functionals ϕn\phi_n. The object that we are after is the Gaussian process with covariance operator Aϕ:S→S′A_\phi: \mathcal{S} \to\mathcal{S}'. Since AϕA_\phi induces the norm of HL′\mathcal{H}'_L (Section 2.2), it is not of trace class, and Sazonov’s theorem [7, 46] rules out a realization of this process in the native space HL\mathcal{H}_L itself. The correct home is the dual of a nuclear space, where the theorem of Minlos [15, 37] converts the continuity of a characteristic functional into a bona fide probability measure. This is the same (S,S′)(\mathcal{S},\mathcal{S}') sandwich that hosts the native space [59, 60], which is what makes the two constructions superposable.

Our contributions are the following.

  1. Generalized Gaussian processes on Hilbert–Gelfand spaces (§ 3.2). Our starting point is the classical result of Gelfand and Vilenkin (Theorem 3.2), which attaches a GGP G∼N(μ,Q)G \sim\mathcal{N}(\mu,Q) to any symmetric positive-semi-definite covariance operator Q:S→S′Q:S \to S'. In Theorem 3.3, we extend the action of the process to test functions in the dual HL′⊃S\mathcal{H}'_L \supset S of the native space, and we characterize the “degenerate” class that corresponds to a pp-conditionally positive-definite covariance: such a process vanishes almost surely on the span Nϕ\mathcal{N}_{\phi} of the boundary functionals, which is the probabilistic transcription of the boundary conditions.

  1. Innovation models (§ 3.3). We show that the processes of this class are exactly those that admit the innovation representation G=Lϕ−1{W}G = L_{\phi}^{-1}\{W\} in terms of a Gaussian white noise WW, or equivalently, that solve the stochastic differential equation LG=WLG = W subject to ϕ(G)=0\phi(G) = 0. The formalism subsumes the classical integral representations: the fractional Brownian motion of Mandelbrot and Van Ness [34] is recovered as the case L=DγL = D^{\gamma} with a single boundary condition, and its higher-order extensions are worked out in Appendix § 7.

  1. Congruence (Proposition 3.5). The Gaussian space of the process—the closed linear span of the random variables G(φ)G(\varphi)—is isometric to the core HA=L∗(L2)\mathcal{H}_{A} = L^{*}(L_{2}), which is the part of the dual HL′\mathcal{H}'_{L} that does not depend on the boundary conditions. Under this isometry, the Riesz pairing of the native space is the covariance of the process, and the innovation provides an orthonormal coordinate system. This is the statement that welds the deterministic and the stochastic descriptions together; Table 1 lists the correspondence term by term.

Variational theoryThis paperRole
LL (pp-admissible)LLwhitening operator
A=(L∗L)−1A = (L^{*}L)^{-1}AAConditional inverse of (L∗L)(L^{*}L)
Aϕ=Pϕ∗APϕA_{\phi} = P_{\phi}^{*}AP_{\phi}CL,ϕC_{L,\phi}covariance operator
A+=Aϕ+RpA_{+} = A_{\phi} + R_{p}A+A_{+}augmented (proper) covariance
Lϕ−1=AϕL∗L_{\phi}^{-1} = A_{\phi}L^{*}shaping operator TTsynthesis of GG from WW
Aϕ=Lϕ−1Lϕ−∗A_{\phi} = L_{\phi}^{-1}L_{\phi}^{-*}CL,ϕ=TT∗C_{L,\phi} = TT^{*}innovation model
ϕ(f)=0\phi(f) = \mathbf{0}ϕ(G)=0\phi(G) = \mathbf{0}boundary conditions
HL\mathcal{H}_{L} (native space)home of gλg_{\lambda}MMSE estimate
HL′\mathcal{H}_{L}^{\prime} (dual)home of φ\varphi and νm\nu_{m}test/measurement functionals
HA=L∗(L2)\mathcal{H}_{A} = L^{*}(L_{2})HA\mathcal{H}_{A}core, congruent to GG\mathcal{G}_{G}
HL,ϕ\mathcal{H}_{L,\phi}HL,ϕ\mathcal{H}_{L,\phi}primary native space
⟨ν,υ⟩HL′\langle\nu,\upsilon\rangle_{\mathcal{H}_{L}^{\prime}}Cov⁡(G(ν),G(υ))\operatorname{Cov}(G(\nu),G(\upsilon))Riesz pairing = covariance
aϕ(x,y)a_{\phi}(x,y)E{G(x)G(y)}\mathbb{E}\{G(x)G(y)\}reproducing kernel = covariance
L2L_{2} (via LL)innovation WWorthonormal coordinates

Table 1. Dictionary between the deterministic objects of the variational theory [59, 60] and their stochastic counterparts in the present paper.

  1. Reproducing kernels and covariance functions (§ 4). For the classic setting where HL\mathcal{H}_{L} is an RKHS on Rd\mathbb{R}^{d}, we identify in Theorem 4.2 a single bivariate function aϕa_{\phi} as the Schwartz kernel of the Gram operator, the reproducing kernel of the primary native space HL,ϕ\mathcal{H}_{L,\phi} (the functions of HL\mathcal{H}_{L} that satisfy the boundary conditions), and the covariance function of the process. The formula involves the Green’s function of the normal operator L∗LL^{*}L, corrected in each variable by the finite-rank term that carries the boundary conditions. (For a trivial null space, the correction vanishes and aϕa_{\phi} reduces to the Green’s function itself, which is the classical kernel [60 Section 4.2].) In Theorem 4.4, we match the operator-level notion of (pp-conditional) positive

definiteness with the kernel-level one of the radial-basis-function literature [8, 66]. As an illustration, we give in § 4.3 a complete and explicit characterization of the isotropic fractional Brownian fields of arbitrary order on Rd\mathbb{R}^{d}—the innovation kernel, the stochastic integral representation à la Mandelbrot, and the pointwise covariance—which appears to be new beyond the classical range H∈(0,1)H \in(0,1) of the Hurst exponent, the range of Lévy’s field and of its fractional extensions.

References

  1. [1]N. Aronszajn. Theory of reproducing kernels. Transactions of the American Mathematical Society, 68(3):337–404, 1950.
  2. [2]A. Ben-Israel. Generalized Inverses. Springer eBook Collection. Springer New York, New York, NY, second edition edition, 2003.
  3. [3]A. Berlinet and C. Thomas-Agnan. Reproducing Kernel Hilbert Spaces in Probability and Statistics, volume 3. Kluwer Academic Boston, 2004.
  4. [4]H. Biermé, O. Durieu, and Y. Wang. Generalized random fields and Lévy’s continuity theorem on the space of tempered distributions. Communications on Stochastic Analysis, 12(4):427–445, 2018.
  5. [5]C. M. Bishop. Pattern Recognition and Machine Learning. Springer, 2006.
  6. [6]T. Blu and M. Unser. Self-similarity: Part II—Optimal estimation of fractal processes. IEEE Transactions on Signal Processing, 55(4):1364–1378, April 2007.DOI
  7. [7]V. I. Bogachev. Gaussian Measures, volume 62 of Mathematical Surveys and Monographs. American Mathematical Society, Providence, RI, 1998.
  8. [8]M. D. Buhmann. Radial Basis Functions: Theory and Implementations, volume 12 of Cambridge Monographs on Applied and Computational Mathematics. Cambridge University Press, Cambridge, 2003.
  9. [9]R. C. Dalang and T. Humeau. Lévy processes and Lévy white noise as tempered distributions. The Annals of Probability, 45(6B):4389–4418, 2017.arxiv.org/abs/1509.05274
  10. [10]E. Di Nezza, G. Palatucci, and E. Valdinoci. Hitchhiker’s guide to the fractional Sobolev spaces. Bulletin des Sciences Mathématiques, 136(5):521–573, 2012.
  11. [11]R. L. Dobrushin. Gaussian and their subordinated self-similar random generalized fields. Annals of Probability, 7(1):1–28, 1979.DOI
  12. [12]J. Duchon. Splines minimizing rotation-invariant semi-norms in Sobolev spaces. In W. Schempp and K. Zeller, editors, Constructive Theory of Functions of Several Variables, volume 571 of Lecture Notes in Mathematics, pages 85–100. Springer, Berlin, 1977.DOI
  13. [13]G. E. Fasshauer and M. J. McCourt. Kernel-based Approximation Methods using MATLAB, volume 19 of Interdisciplinary Mathematical Sciences. World Scientific, Singapore, 2015.DOI
  14. [15]I. M. Gelfand and N. Y. Vilenkin. Generalized Functions. Vol. 4: Applications of Harmonic Analysis. Academic Press, New York, 1964. Translated from the 1961 Russian original.
  15. [17]A. Grothendieck. Produits tensoriels topologiques et espaces nucléaires. Memoirs of the American Mathematical Society, 16:1–140, 1955.DOI
  16. [18]A. E. Hoerl and R. W. Kennard. Ridge regression: Biased estimation for nonorthogonal problems. Technometrics, 12(1):55–67, 1970.
  17. [19]T. Hofmann, B. Schölkopf, and A. J. Smola. Kernel methods in machine learning. Annals of Statistics, 36(3):1171–1220, 2008.
  18. [20]A. Iske. Approximation Theory and Algorithms for Data Analysis, volume 68 of Texts in Applied Mathematics. Springer, Cham, 2018.DOI
  19. [21]S. Janson. Gaussian Hilbert Spaces, volume 129 of Cambridge Tracts in Mathematics. Cambridge University Press, Cambridge, UK, 1997.DOI
  20. [22]T. Kailath. The innovations approach to detection and estimation theory. Proceedings of the IEEE, 58(5):680–695, May 1970.DOI
  21. [23]T. Kailath, R. Geesey, and H. Weinert. Some relations among RKHS norms, Fredholm equations, and innovations representations. IEEE Transactions on Information Theory, 18(3):341–348, May 1972.DOI
  22. [24]T. Kailath, A. H. Sayed, and B. Hassibi. Linear Estimation. Prentice Hall, Upper Saddle River, NJ, 2000.
  23. [25]M. Kanagawa, P. Hennig, D. Sejdinovic, and B. K. Sriperumbudur. Gaussian processes and kernel methods: A review on connections and equivalences. arXiv preprint arXiv:1807.02582, 2018.arxiv.org/abs/1807.02582
  24. [26]G. Kimeldorf and G. Wahba. A correspondence between Bayesian estimation on stochastic processes and smoothing by splines. The Annals of Mathematical Statistics, 41(2):495–502, 1970.DOI
  25. [27]G. Kimeldorf and G. Wahba. Some results on Tchebycheffian spline functions. Journal of Mathematical Analysis and Applications, 33(1):82–95, 1971.DOI
  26. [28]M. Kwaśnicki. Ten equivalent definitions of the fractional Laplace operator. Fractional Calculus and Applied Analysis, 20(1):7–51, 2017.DOI
  27. [29]P. Lévy. Processus stochastiques et mouvement brownien. Gauthier-Villars, Paris, 1948.
  28. [30]F. Lindgren, D. Bolin, and H. Rue. The SPDE approach for Gaussian and non-Gaussian fields: 10 years and still running. Spatial Statistics, 50:100599, 2022.DOI
  29. [31]F. Lindgren, H. Rue, and J. Lindström. An explicit link between Gaussian fields and Gaussian Markov random fields: The stochastic partial differential equation approach. Journal of the Royal Statistical Society: Series B, 73(4):423–498, 2011.DOI
  30. [32]A. Lodhia, S. Sheffield, X. Sun, and S. S. Watson. Fractional Gaussian fields: A survey. Probability Surveys, 13:1–56, 2016.arxiv.org/abs/1407.5598
  31. [33]M. N. Lukić and J. H. Beder. Stochastic processes with sample paths in reproducing kernel Hilbert spaces. Transactions of the American Mathematical Society, 353(10):3945–3969, 2001.DOI
  32. [34]B. B. Mandelbrot and J. W. Van Ness. Fractional Brownian motions, fractional noises and applications. SIAM Review, 10(4):422–437, 1968.DOI
  33. [35]G. Matheron. The intrinsic random functions and their applications. Advances in Applied Probability, 5(3):439–468, 1973.DOI
  34. [36]C. A. Micchelli. Interpolation of scattered data: Distance matrices and conditionally positive definite functions. Constructive Approximation, 2(1):11–22, 1986.
  35. [37]R. A. Minlos. Generalized random processes and their extension to a measure. Trudy Moskovskogo Matematicheskogo Obshchestva, 8:497–518, 1959.
  36. [38]K. Muandet, K. Fukumizu, B. Sriperumbudur, and B. Schölkopf. Kernel mean embedding of distributions: A review and beyond. Foundations and Trends in Machine Learning, 10(1–2):1–141, 2017.DOI
  37. [39]D. E. Myers. Kriging, cokriging, radial basis functions and the role of positive definiteness. Computers & Mathematics with Applications, 24(12):139–148, 1992.DOI
  38. [40]H. Owhadi and C. Scovel. Operator-Adapted Wavelets, Fast Solvers, and Numerical Homogenization: From a Game Theoretic Approach to Numerical Approximation and Algorithm Design, volume 35 of Cambridge Monographs on Applied and Computational Mathematics. Cambridge University Press, Cambridge, UK, 2019.
  39. [41]E. Parzen. An approach to time series analysis. The Annals of Mathematical Statistics, 32(4):951–989, December 1961.DOI
  40. [42]V. I. Paulsen and M. Raghupathi. An Introduction to the Theory of Reproducing Kernel Hilbert Spaces, volume 152 of Cambridge Studies in Advanced Mathematics. Cambridge University Press, Cambridge, 2016.DOI
  41. [43]E. Perrin, R. Harba, C. Berzin-Joseph, I. Iribarren, and A. Bonami. nth-order fractional Brownian motion and fractional Gaussian noises. IEEE Transactions on Signal Processing, 49(5):1049–1059, 2001.DOI
  42. [44]C. E. Rasmussen and C. K. I. Williams. Gaussian Processes for Machine Learning. MIT Press, Cambridge, MA, 2006.DOI
  43. [45]S. Saitoh and Y. Sawano. Theory of Reproducing Kernels and Applications, volume 44 of Developments in Mathematics. Springer, Singapore, 2016.DOI
  44. [46]V. V. Sazonov. A remark on characteristic functionals. Theory of Probability and Its Applications, 3(2):188–192, 1958.DOI
  45. [47]R. Schaback. Native Hilbert spaces for radial basis functions I. In M. D. Buhmann, D. H. Mache, M. Felten, and M. W. Müller, editors, New Developments in Approximation Theory, volume 132 of International Series of Numerical Mathematics, pages 255–282. Birkhäuser, Basel, 1999.DOI
  46. [48]R. Schaback and H. Wendland. Kernel techniques: From machine learning to meshless methods. Acta Numerica, 15:543–639, 2006.DOI
  47. [49]B. Schölkopf, R. Herbrich, and A. J. Smola. A generalized representer theorem. In D. Helmbold and B. Williamson, editors, Computational Learning Theory, pages 416–426. Springer Berlin Heidelberg, 2001.DOI
  48. [50]B. W. Silverman. Some aspects of the spline smoothing approach to non-parametric regression curve fitting. Journal of the Royal Statistical Society. Series B (Methodological), 47(1):1–52, 1985.DOI
  49. [51]T. Sottinen and L. Viitasaari. Transfer principle for nth order fractional Brownian motion with applications to prediction and equivalence in law. Theory of Probability and Mathematical Statistics, 98:199–216, 2019.arxiv.org/abs/1801.07574
  50. [52]I. Steinwart and A. Christmann. Support Vector Machines. Information Science and Statistics. Springer, New York, 2008.
  51. [53]I. Steinwart and C. Scovel. Mercer’s theorem on general domains: On the interaction between measures, kernels, and RKHSs. Constructive Approximation, 35(3):363–417, 2012.DOI
  52. [54]P. D. Tafti and M. Unser. Fractional Brownian vector fields. Multiscale Modeling & Simulation, 8(5):1645–1670, 2010.DOI
  53. [55]A. N. Tikhonov and V. Y. Arsenin. Solution of Ill-posed Problems. Winston & Sons, Washington, 1977.
  54. [56]S. Tirosh, D. Van De Ville, and M. Unser. Polyharmonic smoothing splines and the multidimensional Wiener filtering of fractal-like signals. IEEE Transactions on Image Processing, 15(9):2616–2630, 2006.pubmed.ncbi.nlm.nih.gov/16948307
  55. [57]F. Trèves. Topological Vector Spaces, Distributions and Kernels. Dover Publications, Mineola, NY, 2006. Unabridged republication of the 1967 Academic Press edition.
  56. [58]M. Unser. Sampling—50 years after Shannon. Proceedings of the IEEE, 88(4):569–587, April 2000.DOI
  57. [59]M. Unser. Splines and operators: Part I—admissibility and native spaces. arXiv:2610.14906, 2026.arxiv.org/abs/2610.14906
  58. [60]M. Unser. Splines and operators: Part II—resolution of inverse problems. arXiv:2610.14920, 2026.arxiv.org/abs/2610.14920
  59. [61]M. Unser and T. Blu. Fractional splines and wavelets. SIAM Review, 42(1):43–67, March 2000.DOI
  60. [62]A. W. van der Vaart and J. H. van Zanten. Reproducing kernel Hilbert spaces of Gaussian priors. In Pushing the Limits of Contemporary Statistics: Contributions in Honor of Jayanta K. Ghosh, volume 3 of IMS Collections, pages 200–222. Institute of Mathematical Statistics, Beachwood, OH, 2008.
  61. [63]G. Wahba. Spline Models for Observational Data. Society for Industrial and Applied Mathematics, Philadelphia, 1990.
  62. [64]H. L. Weinert and T. Kailath. Stochastic interpretations and recursive algorithms for spline functions. Annals of Statistics, 2(4):787–794, 1974.DOI
  63. [65]H. L. Weinert and G. S. Sidhu. Stochastic framework for recursive computation of spline functions: Part I, Interpolating splines. IEEE Transactions on Information Theory, 24(1):45–50, 1978.DOI
  64. [66]H. Wendland. Scattered Data Approximation, volume 17 of Cambridge Monographs on Applied and Computational Mathematics. Cambridge University Press, Cambridge, 2005.
  65. [67]N. Wiener. Extrapolation, Interpolation, and Smoothing of Stationary Time Series, with Engineering Applications. MIT Press, Cambridge, MA, 1949.

Paper details

vlnc:W_tst5dhf7
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 7

downloads · 0

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

  • Mathematics

Claude

View statement
Looking for related papers…
Contents