The abstract class Evd<T> in namespace ‘MathNet.Numerics.LinearAlgebra.Factorization’ can get the EigenVectors of a matrix very conveniently.
TestMatrix.Evd().EigenVectors would return the correct result if TestMatrix is a randomly generated matrix with relatively small condition number. But if TestMatrix has a large condition number (>2.0E+13), the result returned is incorrect, while the function ‘eig’ in MATLAB could get the correct result with the same matrix. So I guess the EVD algorithm could be further optimized.

Thank you very much for your contributions!