Conditionally positive definite kernels: theoritical contribution, application to interpolation and approximation

نویسندگان

  • Yves Auffray
  • Pierre Barbillon
چکیده

Since Aronszajn (1950), it is well known that a functional Hilbert space, called Reproducing Kernel Hilbert Space (R.K.H.S ), can be associated to any positive definite kernel K. This correspondance is the basis of many useful algorithms. In the more general context of conditionally positive definite kernels the native spaces are the usual theoretical framework. However, the definition of conditionally positive definite used in that framework is not adapted to extend the results of the positive definite case. We propose a more natural and general definition from which we state a full generalization of Aronszajn’s theorem. It states that for every couple (K,P) such that P is a finite-dimensional vector space of functions and K is a P-conditionally definite positive kernel, there is a unique functional semi-Hilbert space of functions HK,P satisfying a generalized reproducing property. Eventually, we verify that this tool, as native spaces, leads to the same interpolation operator than the one provided by the kriging method and that, using representer theorem, we can identify the solution of a regularized regression problem in HK,P. Key-words: (Conditionally) Positive Definite Kernel, R.K.H.S, Native Space, Interpolation, Kriging, Regularized Regression ∗ Dassault Aviation, Université Paris-sud 11 † Université Paris-sud 11, INRIA SACLAY équipe SELECT in ria -0 03 59 94 4, v er si on 1 9 Fe b 20 09 Noyaux conditionnellement définis positifs : contribution théorique, application à l’interpolation et à l’approximation Résumé : Il est bien connu, depuis Aronszajn (1950), qu’à tout noyau défini positif K, on peut associer un espace de Hilbert de fonctions, appelé espace de Hilbert à noyau reproduisant associé à K (R.K.H.S ). Dans le cas plus général des noyaux conditionnellement positifs, le cadre théorique habituellement invoqué sont les espaces natifs. Cependant, la définition de conditionnellement defini positif qui y est proposée est trop restrictive pour généraliser complétement le cas défini positif. Nous proposons une définition à la fois plus naturelle et plus générale grâce à laquelle une véritable généralisation du théorème d’Aronszajn est démontrée. En substance, il établit qu’à chaque couple (K,P) tel que P est un espace vectoriel de fonctions de dimension finie et K est un noyau P-conditionnellement défini positif, il existe un unique espace semi-Hilbertien de fonctions HK,P (R.K.S.H.S ) satisfaisant une propriété de reproduction généralisée. Nous vérifions que cet outil, comme les espaces natifs, conduit au même opérateur d’interpolation que la méthode du krigeage et que, utilisant le théorème du représentant, on peut identifier la solution d’un problème de régression régularisée dans un R.K.S.H.S . Mots-clés : Noyau (conditionnellement) défini positif, R.K.H.S , espace natif, interpolation, krigeage, régression régularisée in ria -0 03 59 94 4, v er si on 1 9 Fe b 20 09 Conditionally positive definite kernels 3

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

منابع مشابه

Conditionally Positive Definite Kernels and Pontryagin Spaces

Conditionally positive definite kernels provide a powerful tool for scattered data approximation. Many nice properties of such methods follow from an underlying reproducing kernel structure. While the connection between positive definite kernels and reproducing kernel Hilbert spaces is well understood, the analog relation between conditionally positive definite kernels and reproducing kernel Po...

متن کامل

Localized Bases for Kernel Spaces on the Unit Sphere

Approximation/interpolation from spaces of positive definite or conditionally positive definite kernels is an increasingly popular tool for the analysis and synthesis of scattered data and is central to many meshless methods. For a set of N scattered sites, the standard basis for such a space utilizes N globally supported kernels; computing with it is prohibitively expensive for large N . Easil...

متن کامل

Armin Iske * Scattered Data Approximation by Positive Definite Kernel Functions

Kernel functions are suitable tools for scattered data interpolation and approximation. We first review basic features of kernel-based multivariate interpolation, before we turn to the construction and the characterization of positive definite kernels and their associated reproducing kernel Hilbert spaces. The optimality of the resulting kernel-based interpolation scheme is shown. Moreover, we ...

متن کامل

Multivariate Interpolation and Approximation by Translates of a Basis Function

This contribution will touch the following topics: Short introduction into the theory of multivariate interpolation and approximation by nitely many (irregular) translates of a (not necessarily radial) basis function, motivated by optimal recovery of functions from discrete samples. Native spaces of functions associated to conditionally positive definite functions, and relations between such sp...

متن کامل

A trade-off principle in connection with the approximation by positive definite kernels

A frequently used method to handle scattered data approximation problems on different structures is the interpolation by linear combinations of a single positive definite kernel. Here the underlying positive definite kernel is responsible for the quality of the common estimates of the two main issues in this approximation process. Firstly the approximation error and secondly the stability of th...

متن کامل

ذخیره در منابع من


  با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

عنوان ژورنال:

دوره   شماره 

صفحات  -

تاریخ انتشار 2009