lp:~simon-funke/dolfin-adjoint/multistage-optimisation

Branch merges

Propose for merging

No branches dependent on this one.

Related source package recipes

No recipes using this branch. (?)

Create packaging recipe

Related snap packages

Related bugs

Related blueprints

542. By Simon Funke on 2013-03-12

merge from trunk

541. By Simon Funke on 2013-03-12

with the newest dolfin fixes, we can also use more than 2 levels

540. By Simon Funke on 2013-02-24

add a multigrid example that first solves the opt problem on the coarse mesh, projects the solution on the fine mesh and uses that as a starting guess for the optimistion problem on the fine mesh.
Even this simple approach is almost 4 times faster that solving the problem in the fine mesh only

539. By Simon Funke on 2013-02-23

Fix a bug in the steepest descent method where the incorrect line search step values were printed

538. By Simon Funke on 2013-02-23

some fixes to make the bilinear equation run with mutligrid settings

537. By Simon Funke on 2013-02-23

merge

536. By Simon Funke on 2013-02-21

an optimistation example with the navier stokes equations and multigrid

535. By Simon Funke on 2013-02-21

start working on a generic mg opt implementation - not working yet

534. By Simon Funke on 2013-02-21

Some tidying up

533. By Simon Funke on 2013-02-21

add an optional BC argument to the reduced functional object. These BC conditions are imposed
on all gradient and Hessian evaluations. This is useful in cases where the derivative at the boundary is known (e.g. its zero)
but due to numerical errors the computed gradient is not. Without enforcing the correct BC, these errors can add up in particular
in PDE constrained optimisation and can lead to significant errors at the boundary.