How to impose a lower bound to the smallest singular value?

300 Views Asked by At

Is there a simple transformation to apply to a matrix $A$ so that its smallest singular value becomes larger than an arbitrary value $\epsilon$, while the other singular values are perturbed as little as possible?

1

There are 1 best solutions below

0
On

For example: $A+(\sigma_{n-1}-\sigma_n)u_nv_n^T$ will have the same SVD as $A$ except with $\sigma_{n}=\sigma_{n-1}$.