Norme matricielle

Résumé

En mathématiques, une norme matricielle est un cas particulier de norme vectorielle, sur un espace de matrices. Dans ce qui suit, K désigne le corps des réels ou des complexes. Définition Certains auteurs définissent une norme matricielle comme étant simplement une norme sur un espace vectoriel M(K) de matrices à m lignes et n colonnes à coefficients dans K. Pour d'autres, une norme matricielle est seulement définie sur une algèbre M(K) de matrices carrées et est une norme d'algèbre, c'est-à-dire qu'elle est de plus sous-multiplicative. Exemples de normes matricielles Norme de Frobenius La norme de Frobenius sur \mathrm M_{m,n}(K) est celle qui dérive du produit scalaire ou hermitien standard sur cet espace, à savoir (A,B)\in\mathrm M_{m,n}(K)^2\mapsto \langle A,B\rangle =\operatorname{tr}(A^B) =\operatorname{tr}(BA^) , où A^* désigne la matrice adjointe de A et \

Source officielle

https://fr.wikipedia.org/wiki/Norme_matricielle

À propos de ce résultat

Cette page est générée automatiquement et peut contenir des informations qui ne sont pas correctes, complètes, à jour ou pertinentes par rapport à votre recherche. Il en va de même pour toutes les autres pages de ce site. Veillez à vérifier les informations auprès des sources officielles de l'EPFL.

Publications associées (21)

Publications associées

Chargement

Personnes associées (2)

Personnes associées

Chargement

Unités associées (2)

Unités associées

Chargement

Concepts associés (28)

Concepts associés

Chargement

Cours associés (18)

Cours associés

Chargement

Séances de cours associées (32)

Afficher plus

Séances de cours associées

Chargement

Small subset sums

Gergely Ambrus

Let parallel to.parallel to be a norm in R-d whose unit ball is B. Assume that V subset of B is a finite set of cardinality n, with Sigma(v is an element of V) v = 0. We show that for every integer k with 0

Elsevier Science Inc2016

Chargement

Energy-aware adaptive bi-Lipschitz embeddings

Bubacarr Bah, Volkan Cevher, Ali Sadeghian

We propose a dimensionality reducing matrix design based on training data with constraints on its Frobenius norm and number of rows. Our design criteria is aimed at preserving the distances between the data points in the dimensionality reduced space as much as possible relative to their distances in original data space. This approach can be considered as a deterministic Bi-Lipschitz embedding of the data points. We introduce a scalable learning algorithm, dubbed AMUSE, and provide a rigorous estimation guarantee by leveraging game theoretic tools. We also provide a generalization characterization of our matrix based on our sample data. We use compressive sensing problems as an example application of our problem, where the Frobenius norm design constraint translates into the sensing energy.

2013

Chargement

Fast deterministic and randomized algorithms for low-rank approximation, matrix functions, and trace estimation

Alice Cortinovis

In this thesis we propose and analyze algorithms for some numerical linear algebra tasks: finding low-rank approximations of matrices, computing matrix functions, and estimating the trace of matrices.In the first part, we consider algorithms for building low-rank approximations of a matrix from some rows or columns of the matrix itself. We prove a priori error bounds for a greedy algorithm for cross approximation and we develop a faster and more efficient variant of an existing algorithm for column subset selection. Moreover, we present a new deterministic polynomial-time algorithm that gives a cross approximation which is quasi-optimal in the Frobenius norm. The second part of the thesis is concerned with matrix functions. We develop a divide-and-conquer algorithm for computing functions of matrices that are banded, hierarchically semiseparable, or have some other off-diagonal low-rank structure. An important building block of our approach is an existing algorithm for updating the function of a matrix that undergoes a low-rank modification (update), for which we present new convergence results. The convergence analysis of our divide-and-conquer algorithm is related to polynomial or rational approximation of the function.In the third part we consider the problem of approximating the trace of a matrix which is available indirectly, through matrix-vector multiplications. We analyze a stochastic algorithm, the Hutchinson trace estimator, for which we prove tail bounds for symmetric (indefinite) matrices. Then we apply our results to the computation of the (log)determinants of symmetric positive definite matrices.

EPFL2022

Chargement