19 Nov
2015
19 Nov
'15
9:31 a.m.
On 18/11/15 21:43, George Ovchinnikov wrote:
Hi Lawrence,
sqrt(dot(grad(w), grad(w)))
Yes, this is exactly what I meant. I needed to use in Tikhonov regularization in method which I'm working with. I'm updating on every step w based on the rule H(w) - \alpha |\grad(w)|,
where H(w) is basic updating rule, and |\grad(w)| is a penalty for local non-smoothness.
So if w is a Function, you can just add the term alpha * sqrt(dot(grad(w), grad(w))) to your residual and things should work. If you really need it in a different function space you'll have to solve an equation for the mass matrix (to L2 project it into the space). Lawrence