| 1 |
"""Unit test for the PETSc TAO solver""" |
| 2 |
|
| 3 |
# Copyright (C) 2014 Tianyi Li |
| 4 |
# |
| 5 |
# This file is part of DOLFIN. |
| 6 |
# |
| 7 |
# DOLFIN is free software: you can redistribute it and/or modify |
| 8 |
# it under the terms of the GNU Lesser General Public License as published by |
| 9 |
# the Free Software Foundation, either version 3 of the License, or |
| 10 |
# (at your option) any later version. |
| 11 |
# |
| 12 |
# DOLFIN is distributed in the hope that it will be useful, |
| 13 |
# but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 14 |
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
| 15 |
# GNU Lesser General Public License for more details. |
| 16 |
# |
| 17 |
# You should have received a copy of the GNU Lesser General Public License |
| 18 |
# along with DOLFIN. If not, see <http://www.gnu.org/licenses/>. |
| 19 |
# |
| 20 |
# First added: 2014-07-19 |
| 21 |
# Last changed: 2014-10-14 |
| 22 |
|
| 23 |
"""This demo uses PETSc TAO solver for nonlinear (bound-constrained) |
| 24 |
optimisation problems to solve the same minimization problem for testing |
| 25 |
TAOLinearBoundSolver.""" |
| 26 |
|
| 27 |
from dolfin import * |
| 28 |
import pytest |
| 29 |
|
| 30 |
from dolfin_utils.test import * |
| 31 |
|
| 32 |
backend = set_parameters_fixture("linear_algebra_backend", ["PETSc"]) |
| 33 |
|
| 34 |
@skip_if_not_PETSc |
| 35 |
def test_tao_linear_bound_solver(backend): |
| 36 |
"Test PETScTAOSolver" |
| 37 |
|
| 38 |
# Create mesh and define function space |
| 39 |
Lx = 1.0; Ly = 0.1 |
| 40 |
mesh = RectangleMesh(Point(0, 0), Point(Lx, Ly), 100, 10) |
| 41 |
V = FunctionSpace(mesh, "Lagrange", 1) |
| 42 |
|
| 43 |
# Define Dirichlet boundaries |
| 44 |
def left(x,on_boundary): |
| 45 |
return on_boundary and near(x[0], 0.0) |
| 46 |
|
| 47 |
def right(x,on_boundary): |
| 48 |
return on_boundary and near(x[0], Lx) |
| 49 |
|
| 50 |
# Define boundary conditions |
| 51 |
zero = Constant(0.0) |
| 52 |
one = Constant(1.0) |
| 53 |
bc_l = DirichletBC(V, zero, left) |
| 54 |
bc_r = DirichletBC(V, one, right) |
| 55 |
bcs = [bc_l, bc_r] |
| 56 |
|
| 57 |
# Define the variational problem |
| 58 |
u = Function(V) |
| 59 |
du = TrialFunction(V) |
| 60 |
v = TestFunction(V) |
| 61 |
cv = Constant(3.0/4.0) |
| 62 |
ell = Constant(0.5) # This should be smaller than Lx |
| 63 |
energy = cv*(ell/2.0*inner(grad(u), grad(u))*dx + 2.0/ell*u*dx) |
| 64 |
grad_energy = derivative(energy, u, v) |
| 65 |
H_energy = derivative(grad_energy, u, du) |
| 66 |
|
| 67 |
# Define the lower and upper bounds |
| 68 |
lb = interpolate(Constant(0.0), V) |
| 69 |
ub = interpolate(Constant(1.0), V) |
| 70 |
|
| 71 |
# Apply BC to the lower and upper bounds |
| 72 |
for bc in bcs: |
| 73 |
bc.apply(lb.vector()) |
| 74 |
bc.apply(ub.vector()) |
| 75 |
|
| 76 |
# Define the minimisation problem by using OptimisationProblem class |
| 77 |
class TestProblem(OptimisationProblem): |
| 78 |
|
| 79 |
def __init__(self): |
| 80 |
OptimisationProblem.__init__(self) |
| 81 |
|
| 82 |
# Objective function |
| 83 |
def f(self, x): |
| 84 |
u.vector()[:] = x |
| 85 |
return assemble(energy) |
| 86 |
|
| 87 |
# Gradient of the objective function |
| 88 |
def F(self, b, x): |
| 89 |
u.vector()[:] = x |
| 90 |
assemble(grad_energy, tensor=b) |
| 91 |
|
| 92 |
# Hessian of the objective function |
| 93 |
def J(self, A, x): |
| 94 |
u.vector()[:] = x |
| 95 |
assemble(H_energy, tensor=A) |
| 96 |
|
| 97 |
# Create the PETScTAOSolver |
| 98 |
solver = PETScTAOSolver() |
| 99 |
|
| 100 |
# Set some parameters |
| 101 |
solver.parameters["monitor_convergence"] = True |
| 102 |
solver.parameters["report"] = True |
| 103 |
|
| 104 |
solver.solve(TestProblem(), u.vector(), lb.vector(), ub.vector()) |
| 105 |
solver.solve(TestProblem(), u.vector(), lb.vector(), ub.vector()) |
| 106 |
|
| 107 |
# Verify that energy(u) = Ly |
| 108 |
assert round(assemble(energy) - Ly, 4) == 0 |