Accurate SVDs of polynomial Vandermonde matrices involving orthonormal polynomials (Q2502887): Difference between revisions

A polynomial Vandermonde matrix involving orthonormal polynomials is a matrix of the form: \(V_P(x)=[P_{i-1}(x_j)]_{i,j=1}^n\) where \(P_i\) are real polynomials, orthonormal on some interval \([a,b],\) \(\deg P_i=i,\) and \(x=(x_1,x_2,\dots ,x_n)\) is a complex \(n\)-vector. The authors develop a new algorithm for computing the singular values decomposition of \(V_P(x)\) which is more accurate and efficient than the traditional ones and is independent of the condition number. The authors show that the singular values of \(V_P(x)\) are well conditioned functions of the data. They demonstrate the efficiency of their algorithm with numerical examples. They also explain how this theory can be extended to other polynomial Vandermonde matrices, involving polynomials that are not orthonormal or even orthogonal.