problem with vector-valued Constant control

hello,

i slightly adapted an example which i found for the use of Constant controls, so that it uses vector valued Constants (the commented code uses 3 scalar valued Constant controls and works without error):

from dolfin import *
from dolfin_adjoint import *

mesh = UnitCubeMesh(2,2,2)

V = FunctionSpace(mesh, "CG", 1)

u = Function(V)
v = TestFunction(V)

f = Constant((1.,2.,3.))
#f1 = Constant(1.)
#f2 = Constant(1.)
#f3 = Constant(1.)

#F = inner(grad(u), grad(v))*dx - inner(as_vector([f1,f2,f3]), grad(v))*dx
F = inner(grad(u), grad(v))*dx - inner(f, grad(v))*dx
solve(F == 0, u)

functional = Functional(u * dx)

reduced_functional = ReducedFunctional(functional, Control(f))
#reduced_functional = ReducedFunctional(functional, [Control(f1), Control(f2), Control(f3)])
problem = MinimizationProblem(reduced_functional)

opt = minimize(reduced_functional)

The forward problem seems to work, but unfortunately the optimization fails with the following error:

No Jacobian form specified for nonlinear variational problem.
Differentiating residual form F to obtain Jacobian J = F'.
Solving nonlinear variational problem.
  Newton iteration 0: r (abs) = 2.016e+00 (tol = 1.000e-10) r (rel) = 1.000e+00 (tol = 1.000e-09)
  Newton iteration 1: r (abs) = 1.516e-15 (tol = 1.000e-10) r (rel) = 7.521e-16 (tol = 1.000e-09)
  Newton solver finished in 1 iterations and 1 linear solver iterations.
RUNNING THE L-BFGS-B CODE

           * * *

Machine precision = 2.220D-16
 N =            3     M =           10
 This problem is unconstrained.

At X0         0 variables are exactly at the bounds
Couldn't map 'f_7' to a float, returning ufl object without evaluation.
Solving nonlinear variational problem.
  Newton iteration 0: r (abs) = 2.016e+00 (tol = 1.000e-10) r (rel) = 1.000e+00 (tol = 1.000e-09)
  Newton iteration 1: r (abs) = 1.516e-15 (tol = 1.000e-10) r (rel) = 7.521e-16 (tol = 1.000e-09)
  Newton solver finished in 1 iterations and 1 linear solver iterations.
Coefficient and argument shapes do not match!
Traceback (most recent call last):
  File "constant_control.py", line 26, in <module>
    opt = minimize(reduced_functional)
  File "/usr/lib/python2.7/dist-packages/dolfin_adjoint/misc.py", line 33, in func_wrapper
    res = func(*args, **kwargs)
  File "/usr/lib/python2.7/dist-packages/dolfin_adjoint/optimization/optimization.py", line 223, in minimize
    opt = algorithm(rf_np, **kwargs)
  File "/usr/lib/python2.7/dist-packages/dolfin_adjoint/optimization/optimization.py", line 128, in minimize_scipy_generic
    res = scipy_minimize(J, m_global, method=method, **kwargs)
  File "/usr/lib/python2.7/dist-packages/scipy/optimize/_minimize.py", line 447, in minimize
    callback=callback, **options)
  File "/usr/lib/python2.7/dist-packages/scipy/optimize/lbfgsb.py", line 330, in _minimize_lbfgsb
    f, g = func_and_grad(x)
  File "/usr/lib/python2.7/dist-packages/scipy/optimize/lbfgsb.py", line 279, in func_and_grad
    g = jac(x, *args)
  File "/usr/lib/python2.7/dist-packages/dolfin_adjoint/optimization/optimization.py", line 62, in <lambda>
    dJ = lambda m: rf_np.derivative(m, forget=forget, project=project)
  File "/usr/lib/python2.7/dist-packages/dolfin_adjoint/reduced_functional_numpy.py", line 76, in derivative
    dJdm = self.__base_derivative__(forget=forget, project=project)
  File "/usr/lib/python2.7/dist-packages/dolfin_adjoint/reduced_functional.py", line 255, in derivative
    dfunc_value = drivers.compute_gradient(self.functional, self.controls, forget=forget, project=project)
  File "/usr/lib/python2.7/dist-packages/dolfin_adjoint/drivers.py", line 160, in compute_gradient
    out = param.equation_partial_derivative(adjglobals.adjointer, output.data, i, fwd_var)
  File "/usr/lib/python2.7/dist-packages/dolfin_adjoint/controls.py", line 473, in equation_partial_derivative
    return [p.equation_partial_derivative(adjointer, adjoint, i, variable) for p in self.controls]
  File "/usr/lib/python2.7/dist-packages/dolfin_adjoint/controls.py", line 216, in equation_partial_derivative
    backend.derivative(form, get_constant(self.a), dparam))
  File "/usr/lib/python2.7/dist-packages/dolfin/fem/formmanipulations.py", line 90, in derivative
    return ufl.derivative(form, u, du, coefficient_derivatives)
  File "/usr/lib/python2.7/dist-packages/ufl/formoperators.py", line 283, in derivative
    argument)
  File "/usr/lib/python2.7/dist-packages/ufl/formoperators.py", line 227, in _handle_derivative_arguments
    error("Coefficient and argument shapes do not match!")
  File "/usr/lib/python2.7/dist-packages/ufl/log.py", line 171, in error
    raise self._exception_type(self._format_raw(*message))
ufl.log.UFLException: Coefficient and argument shapes do not match!

Is there a problem with the example or a vector-valued Constant control not supported by dolfin adjoint?

thanks and best wishes

Florian

Comments (1)