Structured matrices are parameterised by the internal representation:
template<class R> struct MatrStr<R>;
Structured matrices are two-dimensional arrays which can be represented by values, where is the size of the matrix. This class provides classical arithmetic operations such as addition, substraction, multiplication by scalars, vectors, matrices . . . The product of two structured matrices is usually not a structured matrices (at least of the same type) and therefore is not implemented. The product of a structured matrices by a vector can usually be performed in almost linear time, ie. in (or according to the field), using FFT.
A matrix is a Toeplitz matrix if for all , the entry depends only on , that is if for all pairs of and for which the entries and are defined. Associated with a Toeplitz matrix, we have the Toeplitz operator which is a projection composed with the multiplication operator by a univariate polynomial in (see [BP94]). An interesting point of this definition is that the product of a Toeplitz matrix by a vector is a subvector of the coefficient vector of the product of two polynomials of K[x]. This product can be done within (or according to the field) operations using FFT.
An implementation of such data-structure is available as
- See also:
synaps/linalg/toeplitz.h
A matrix is a Hankel matrix if its entries depends only on , that is, if for all pairs for which the entries are defined. To this kind of matrix, we can associate a Hankel operator which is defined as the projection of the multiplication by a fixed Laurent polynomial (a Laurent polynomial is a polynomial of , that is a polynomial in both the variables and ). So we can compute the product of a Hankel matrix by a vector as a subvector of the product of a fixed polynomial by a polynomial in . This can also be done within ops (or according to the field), using FFT. See [4].
An implementation of such data-structure is available as
- See also:
synaps/linalg/hankel.h
#include <synaps/linalg/MatrStr.H>
#include <synaps/linalg/hankel.H>
typedef MatrStr<linalg::hankel<double> > mstr_t;
int main(int argc, char** argv)
{
double v[]={1,2,1,1,3,4,2,2,5,5,5,5,-2, 1,2,3,1, -2.8,-2.4,1,.2,5.8};
mstr_t A(4,4,v);
std::cout << A << std::endl;
- See also:
synaps/linalg/MatrStr.h