Does exchanging the role data fit term and regularization term make sens?

36 Views Asked by At

For example the total variation regularization formulation is given as $$ \min_x \frac{1}{2} || Ax -y ||_2^2 + \alpha ||\nabla x||_1. $$

Now I am wondering, would it make sense to ask for a solution of the minimization problem given as $$ \min_x ||\nabla x||_1 + \frac{\alpha}{2} || Ax -y ||_2^2, $$ where now the data fit term acts as a regularization term? Since I have never seen something like this in literature until now, I would guess this is not feasible (for maybe obvious reasons), but why is that? If yes, what are suitable algorithms to solve problems of this form?

Thank you in advance.

1

There are 1 best solutions below

0
On

Under optimization, we have that $$ ||\nabla x||_1 + \frac{\alpha}{2}||Ax-y||^2_2 \equiv \frac{1}{2}||Ax-y||^2_2+\frac{1}{\alpha}||\nabla x||_1, $$ since it is just a scaling by the positive constant $\alpha^{-1}$. Hence your problem is also a TV-problem.