How to get the derivative of the solution with respect to a parameter

dolfin-adjoint
1

I’m trying to implement a stochastic filter for a system based on a time-dependent PDE (heat equation). The PDE depends on an unknown potential that I eventually want to estimate. For this, I need the derivative of the solution at a time t with respect to the potential. How do I get this?
I know that what I am looking for is basically the solution to the tangent linear equation, but I don’t want to implement this by myself, because I assume that dolfin-adjoint is able to do this much better.

from fenics import *
from fenics_adjoint import *

T = 50.0     #time of evaluation
N_t = 200     #number of time steps
dt = T / N_t  #time step size
N_s = 64      #space mesh size

# Set initial value
u_0 = Expression('0', degree=1)
# Set boundary value
u_D = Expression('0', degree=1)

# Basic setup
mesh = UnitIntervalMesh(N_s)
V = FunctionSpace(mesh, 'P', 2)
def boundary(x, on_boundary):
    return on_boundary
bc = DirichletBC(V, u_D, boundary)

# Define the potential
q = interpolate(Expression('0.1-0.05*sin(2*pi*(x[0]-0.25))', degree=1), V)

control = Control(q) # Want to differentiate with respect to q

# Define the variational problem (using backwards Euler)
u_next = Function(V)
u = interpolate(u_0,V)
v = TestFunction(V)
r = Expression('x[0] >= 0.215 && x[0] <= 0.315 ? (4*sin(4*pi*t)+0.001*pow(t,2)) : 0', \
               degree=1, t=0)
F = ( (u_next - u)*v /dt + dot(q*grad(u_next), grad(v)) - r*v )*dx

# Start solution
for t in range(N_t):
    # Update current time
    t += dt
    # At time n, we want r = r((n+1)dt)
    r.t = t
    # Compute solution
    solve(F == 0, u_next, bc)
    # Update u
    u.assign(u_next)

# Now, I want the derivative of u(T) with respect to q
dudq = compute_gradient(u, control)   # But obviously, this does not work.
2

Hi,
Please read through:
http://www.dolfin-adjoint.org/en/latest/documentation/maths/index.html#dolfin-adjoint-mathematical-background
Dolfin-adjoint does not compute the direct du/dm, But the reduced functional. Therefore, you need to define a functional.

3

For anyone interested, see also https://bitbucket.org/dolfin-adjoint/pyadjoint/issues/147/derivative-of-solution-wrt-parameter

4

Okay, I will see if I can find a workaround.

Thank you very much

5

Hi, I am facing the exact same problem (need the derivative of the solution wrt a parameter that appears in the initial PDE). Have you managed to find a way to do it in the end?