I am working with classical multidimensional scaling and the configuration matrix for start the iterative process is ill-conditioned. For that reasen I am looking for methods to transform this matrix and obtain a better condition number

More Damian Valdés Santiago's questions See All
Similar questions and discussions