if i_best != indices_better[-1]:
print(i_best, imod)
indices_better.append(imod)
print(indices_better)
print(len(result.models))
models = [result.models[i] for i in indices_better]
else:
models = None
outputs = result.compute_models(models)
# plot results
if rank == 0:
out_dir = result.meta['out_dir']
points = NeighbouhoodAlgorithm._get_points_for_voronoi(
result.perturbations, range_dict, model_params._types)
points = points[:, model_params.get_free_indices()]
# points = np.array(result.perturbations)[
# :, model_params.get_free_indices()]
misfits = result.misfit_dict['variance'].mean(axis=1)
n_r = result.meta['n_r']
n_s = result.meta['n_s']
n_mod = points.shape[0]
i_out = 0
for imod in range(n_mod):
figpath = os.path.join(
out_dir, 'voronoi_{:05d}.png'.format(imod))
fig = plt.figure(figsize=(17,5))
gs = gridspec.GridSpec(1, 4, width_ratios=[2, 1, 1, 1])
# plot voronoi cells
ax0 = fig.add_subplot(gs[0])
_, _, colormap = plot_voronoi_2d(
dtype='bool')
mask_dict[ParameterType.RADIUS][[0, 1]] = False
equal_dict = dict()
equal_dict[ParameterType.VSH] = np.arange(model_params._n_grd_params,
dtype='int')
equal_dict[ParameterType.VSH][2] = 0
equal_dict[ParameterType.VSH][3] = 1
discon_arr = np.zeros(model_params._n_nodes, dtype='bool')
discon_arr[1] = True
model_params.set_constraints(mask_dict=mask_dict,
equal_dict=equal_dict,
discon_arr=discon_arr)
print('Free indices:', model_params.get_free_indices())
rng = np.random.default_rng(seed=0)
fig, ax = plt.subplots(1)
for i in range(5):
values = np.zeros((model_params._n_grd_params, len(types)))
values[:, 0] = rng.uniform(-0.5, 0.5, size=(values.shape[0]))
values[:, 1] = rng.uniform(-100, 100, size=(values.shape[0]))
values_dict = {
param_type: values[:, i]
for i, param_type in enumerate(types)
}
model_gen = model.build_model(model, model_params, values_dict)
model_gen.plot(types=[ParameterType.VSH], ax=ax)
plt.show()