Added solver options and revised plotting.

examples/nonconvex/tidalfarm/README.md

@@ -5,9 +5,9 @@ using the conditional gradient method.
 
 ## Graphical illustrations
 
-![](output/26-January-2023-12-12-15_solution_best_N=100_online_version.png)
+![](output/10-May-2023-13-44-27_solution_best_n=100_online_version.png)
 |:--:| 
-*Optimal turbine friction*
+*Optimal turbine density*
 
 Using the conditional gradient method, we can compute a tidal-stream farm layout
 which maximizes energy extraction. The setting used for this application is descriped

examples/nonconvex/tidalfarm/optimize.py

@@ -25,19 +25,19 @@
 set_log_level(30)
 
 import fw4pde
+from solver_options import SolverOptions
 
 state = steady_sw(control)
 u, p = split(state)
 
+# sparsity parameter and box constraints
 WtoMW = 1e-6
-
 beta = WtoMW*4800.0
 lb = Constant(0.0)
 ub = Constant(0.059)
 
+# Objective function
 scaled_L1_norm = fw4pde.problem.ScaledL1Norm(control_space,beta)
-
-# power functional appears to be implemented in https://zenodo.org/record/224251
 power_functional = assemble(rho*control*inner(u,u)**1.5*site_dx(1))
 
 ctrl = Control(control)
@@ -49,22 +49,18 @@
 box_constraints = fw4pde.problem.BoxConstraints(control_space, lb, ub)
 moola_box_lmo = fw4pde.algorithms.MoolaBoxLMO(box_constraints.lb, box_constraints.ub, beta)
 
-stepsize = fw4pde.stepsize.DecreasingStepSize()
-#stepsize = fw4pde.stepsize.DunnScalingStepSize()
-#stepsize = fw4pde.stepsize.DemyanovRubinovOptimalStepSize()
-#stepsize = fw4pde.stepsize.DemyanovRubinovAdaptiveStepSize()
-#stepsize = fw4pde.stepsize.DecreasingAdaptiveStepSize()
 
-gtol= 1e-4
-ftol = -np.inf
-maxiter = 10
-options = {"maxiter": maxiter, "gtol": gtol, "ftol": ftol}
+solver_options = SolverOptions()
+options = solver_options.options
+stepsize = solver_options.stepsize
 
 solver = fw4pde.algorithms.FrankWolfe(problem, initial_point=u_moola, nonsmooth_functional=scaled_L1_norm,\
                 stepsize=stepsize, lmo=moola_box_lmo, options=options)
 
 sol = solver.solve()
 
+# Postprocessing: Plotting and saving
+
 solution_final = sol["control_final"].data
 plt.set_cmap("coolwarm")
 c = plot(solution_final)