Documentation clarification; monitoring objective functional along iterations about proxtv HOT 3 CLOSED

Hi there!

Expressions in the form |x_i - x_{i-1}| indeed refer to absolute values (or l1-norms). This is true for all the problems presented where this notation is used. When other norms are used (l2, lp) you will see that explicitly mentioned. Note also that proxTV can solve general lp variations of any of these problems (with p >= 1), see Generalized N-dimensional Anisotropic Total Variation. However, proxTV is mainly optimized for l1 problems.

Regarding monitoring the value of the objective being minimized, I'm afraid that is not possible directly. I tried to make the code as efficient as possible, and that involved reducing book-keeping and gathered statistics to a minimum. Nevertheless, some of the methods accept a max_iters parameter. For my performance experiments I called the same method several times with different max_iters, to have a look at how close the obtained solution is to optimum (max_iters = some large number). I don't know if this is good enough for you, though.

from proxtv.

from proxtv.

Thanks for proposing the change, but I'd rather keep the code clean of this. Though my code is not something I would call "clean", but I'd rather not introduce additional complexity into it.

from proxtv.