Verification of positive definiteness (Q2502318)

An efficient numerical criterion to verify positive definiteness of a symmetric or Hermitian matrix is presented. The criterion is based on standard IEEE 754 floating-point arithmetic with rounding to nearest. It implements a single floating-point Cholesky decomposition and improves a known result. A major advantage of the method described is that any library routine can be used. Underflows are allowed so that the verification is rigorous, including all possible effects of rounding errors. Using the criterion an efficient algorithm to compute rigorous error bounds for the solution of linear systems with symmetric positive definite matrix is proposed. A computational criterion to verify that a given symmetric or Hermitian matrix is not positive definite is given as well. Several computational examples demonstrate the effectiveness of the criteria presented.