Modify Reduced Functional during optimization

Question

Hi!

Is there a way of modify the reduced functional during an optimization? I am asking because I want to use the scheme of updating coefficients in an multi objective optimization. This way, the parts that compound the multi objective functional do not become very different from each other (numerically).

Thanks!

Answer 1

I believe I found!

For a Functional with respective Reduced Functional:

J = Functional(w_e * Energy_dissipation(u, rho) + \
  (en_coeff(u, rho)/vort_coeff(u, rho)) * w_v * Vorticity(u,rho) + \
  (en_coeff(u, rho)/torq_coeff(u, rho)) * w_t * Torque(u,rho))

  m = Control(rho)                                        # Control
  Jhat = ReducedFunctional(J, m, eval_cb_post = eval_cb)  # Reduced Functional

I am defining the following coefficients:

def en_coeff(u, rho):
    (u_tape,p_tape) = split(DolfinAdjointVariable(w).tape_value())
    return assemble(Energy_dissipation(u_tape, rho))

def vort_coeff(u, rho):
    (u_tape,p_tape) = split(DolfinAdjointVariable(w).tape_value())
    return assemble(Vorticity(u_tape, rho))

def torq_coeff(u, rho):
    (u_tape,p_tape) = split(DolfinAdjointVariable(w).tape_value())
    return assemble(Torque(u_tape, rho))

Code is still a little messy, sorry about that. Could anyone confirm that this really changes during the optimization?

Comments

Forget....it doesn't seem to be working..