Floating-point error mitigationFloating-point error mitigation is the minimization of errors caused by the fact that real numbers cannot, in general, be accurately represented in a fixed space. By definition, floating-point error cannot be eliminated, and, at best, can only be managed. Huberto M. Sierra noted in his 1956 patent "Floating Decimal Point Arithmetic Control Means for Calculator": Thus under some conditions, the major portion of the significant data digits may lie beyond the capacity of the registers.
Factorisation de CholeskyLa factorisation de Cholesky, nommée d'après André-Louis Cholesky, consiste, pour une matrice symétrique définie positive , à déterminer une matrice triangulaire inférieure telle que : . La matrice est en quelque sorte une « racine carrée » de . Cette décomposition permet notamment de calculer la matrice inverse , de calculer le déterminant de A (égal au carré du produit des éléments diagonaux de ) ou encore de simuler une loi multinormale. Elle est aussi utilisée en chimie quantique pour accélérer les calculs (voir Décomposition de Cholesky (chimie quantique)).
Matrix decompositionIn the mathematical discipline of linear algebra, a matrix decomposition or matrix factorization is a factorization of a matrix into a product of matrices. There are many different matrix decompositions; each finds use among a particular class of problems. In numerical analysis, different decompositions are used to implement efficient matrix algorithms. For instance, when solving a system of linear equations , the matrix A can be decomposed via the LU decomposition.
Numerical methods for linear least squaresNumerical methods for linear least squares entails the numerical analysis of linear least squares problems. A general approach to the least squares problem can be described as follows. Suppose that we can find an n by m matrix S such that XS is an orthogonal projection onto the image of X. Then a solution to our minimization problem is given by simply because is exactly a sought for orthogonal projection of onto an image of X (see the picture below and note that as explained in the next section the image of X is just a subspace generated by column vectors of X).
Bfloat16 floating-point formatThe bfloat16 (brain floating point) floating-point format is a computer number format occupying 16 bits in computer memory; it represents a wide dynamic range of numeric values by using a floating radix point. This format is a truncated (16-bit) version of the 32-bit IEEE 754 single-precision floating-point format (binary32) with the intent of accelerating machine learning and near-sensor computing. It preserves the approximate dynamic range of 32-bit floating-point numbers by retaining 8 exponent bits, but supports only an 8-bit precision rather than the 24-bit significand of the binary32 format.
Error analysis (mathematics)In mathematics, error analysis is the study of kind and quantity of error, or uncertainty, that may be present in the solution to a problem. This issue is particularly prominent in applied areas such as numerical analysis and statistics. In numerical simulation or modeling of real systems, error analysis is concerned with the changes in the output of the model as the parameters to the model vary about a mean. For instance, in a system modeled as a function of two variables Error analysis deals with the propagation of the numerical errors in and (around mean values and ) to error in (around a mean ).
Numerical methods for partial differential equationsNumerical methods for partial differential equations is the branch of numerical analysis that studies the numerical solution of partial differential equations (PDEs). In principle, specialized methods for hyperbolic, parabolic or elliptic partial differential equations exist. Finite difference method In this method, functions are represented by their values at certain grid points and derivatives are approximated through differences in these values.
Probabilistic numericsProbabilistic numerics is an active field of study at the intersection of applied mathematics, statistics, and machine learning centering on the concept of uncertainty in computation. In probabilistic numerics, tasks in numerical analysis such as finding numerical solutions for integration, linear algebra, optimization and simulation and differential equations are seen as problems of statistical, probabilistic, or Bayesian inference.
Catastrophic cancellationIn numerical analysis, catastrophic cancellation is the phenomenon that subtracting good approximations to two nearby numbers may yield a very bad approximation to the difference of the original numbers. For example, if there are two studs, one long and the other long, and they are measured with a ruler that is good only to the centimeter, then the approximations could come out to be and . These may be good approximations, in relative error, to the true lengths: the approximations are in error by less than 2% of the true lengths, .
Norme d'opérateurEn mathématiques, et plus particulièrement en analyse fonctionnelle, une norme d'opérateur ou norme subordonnée est une norme définie sur l'espace des opérateurs bornés entre deux espaces vectoriels normés. Entre deux tels espaces, les opérateurs bornés ne sont autres que les applications linéaires continues. Sur un corps K « valué » (au sens : muni d'une valeur absolue) et non discret (typiquement : K = R ou C), soient E et F deux espaces vectoriels normés respectivement munis des normes ‖ ‖ et ‖ ‖.
Matrice inversibleEn mathématiques et plus particulièrement en algèbre linéaire, une matrice inversible (ou régulière ou encore non singulière) est une matrice carrée A pour laquelle il existe une matrice B de même taille n avec laquelle les produits AB et BA sont égaux à la matrice identité. Dans ce cas la matrice B est unique, appelée matrice inverse de A et notée B = A. Cette définition correspond à celle d’élément inversible pour la multiplication dans l’anneau des matrices carrées associé.
Matrice unitaireEn algèbre linéaire, une matrice carrée U à coefficients complexes est dite unitaire si elle vérifie les égalités : où la matrice adjointe de U est notée U* (ou U en physique, et plus particulièrement en mécanique quantique) et I désigne la matrice identité. L'ensemble des matrices unitaires de taille n forme le groupe unitaire U(n). Les matrices unitaires carrées à coefficients réels sont les matrices orthogonales.
