SVD-based approximations of bivariate functions

Author

Bizzarri, Federico ; Parodi, Mauro ; Storace, Marco

Author_Institution

Dept. of Biophys. & Electron. Eng., Univ. of Genoa, Genova, Italy

fYear

2005

fDate

23-26 May 2005

Firstpage

4915

Abstract

A method to approximate functions of two variables is presented; it is suitable for hardware implementations based on digital or mixed signal architectures. Such a method is based on the properties of the singular value decomposition (SVD) of a matrix that stores the samples of the function to be approximated. The considered SVD-based approximations are expressed as products of functions of a single variable and are built up as combinations of proper sets of piecewise-linear or polynomial basis functions. In the proposed examples, the accuracy of the SVD-based approximations is compared with that obtained by resorting to two well-assessed methods.

Keywords

approximation theory; digital integrated circuits; functions; integrated circuit design; mixed analogue-digital integrated circuits; piecewise linear techniques; polynomials; singular value decomposition; SVD-based approximations; bivariate functions; digital or architectures; hardware implementations; mixed signal architectures; piecewise-linear functions; polynomial basis functions; singular value decomposition; Approximation methods; Field programmable gate arrays; Hardware; Matrix decomposition; Neural networks; Nonlinear circuits; Piecewise linear techniques; Polynomials; Singular value decomposition; Support vector machines;

fLanguage

English

Publisher

ieee

Conference_Titel

Circuits and Systems, 2005. ISCAS 2005. IEEE International Symposium on

Print_ISBN

0-7803-8834-8

Type

conf

DOI

10.1109/ISCAS.2005.1465735

Filename

1465735