def test_inv(self):
reg = regularization.Tikhonov(self.mesh)
opt = optimization.InexactGaussNewton(maxIter=10, use_WolfeCurvature=True)
invProb = inverse_problem.BaseInvProblem(self.dmiscombo, reg, opt)
directives_list = [
directives.ScalingMultipleDataMisfits_ByEig(verbose=True),
directives.AlphasSmoothEstimate_ByEig(verbose=True),
directives.BetaEstimate_ByEig(beta0_ratio=1e-2),
directives.MultiTargetMisfits(TriggerSmall=False),
directives.BetaSchedule(),
]
inv = inversion.BaseInversion(invProb, directiveList=directives_list)
m0 = self.model.mean() * np.ones_like(self.model)
mrec = inv.run(m0)
def test_inv_mref_setting(self):
reg1 = regularization.Tikhonov(self.mesh)
reg2 = regularization.Tikhonov(self.mesh)
reg = reg1 + reg2
opt = optimization.ProjectedGNCG(
maxIter=30, lower=-10, upper=10, maxIterLS=20, maxIterCG=50, tolCG=1e-4
)
invProb = inverse_problem.BaseInvProblem(self.dmiscombo, reg, opt)
directives_list = [
directives.ScalingMultipleDataMisfits_ByEig(
chi0_ratio=[0.01, 1.0], verbose=False
),
directives.AlphasSmoothEstimate_ByEig(verbose=False),
directives.BetaEstimate_ByEig(beta0_ratio=1e-2),
directives.MultiTargetMisfits(TriggerSmall=False, verbose=True),
directives.BetaSchedule(),
]
inv = inversion.BaseInversion(invProb, directiveList=directives_list)
m0 = self.model.mean() * np.ones_like(self.model)
mrec = inv.run(m0)
self.assertTrue(np.all(reg1.mref == m0))
self.assertTrue(np.all(reg2.mref == m0))
alpha0_ratio = np.r_[np.zeros(len(reg.objfcts[0].objfcts)),
1e-2 * np.ones(len(reg.objfcts[1].objfcts)),
1e-2 * 100.0 * np.ones(len(reg.objfcts[2].objfcts)), ]
Alphas = directives.AlphasSmoothEstimate_ByEig(alpha0_ratio=alpha0_ratio,
verbose=True)
# initialize beta and beta/alpha_s schedule
beta = directives.BetaEstimate_ByEig(beta0_ratio=1e-4)
betaIt = directives.PGI_BetaAlphaSchedule(
verbose=True,
coolingFactor=2.0,
tolerance=0.2,
progress=0.2,
)
# geophy. and petro. target misfits
targets = directives.MultiTargetMisfits(
verbose=True,
chiSmall=0.5, # ask for twice as much clustering (target value is /2)
)
# add learned mref in smooth once stable
MrefInSmooth = directives.PGI_AddMrefInSmooth(
wait_till_stable=True,
verbose=True,
)
# update the parameters in smallness (L2-approx of PGI)
update_smallness = directives.PGI_UpdateParameters(
update_gmm=True, # update the GMM each iteration
kappa=np.c_[ # confidences in each mean phys. prop. of each cluster
1e10 * np.ones(
2
), # fixed background at 0 density, 0 mag. susc. (high confidences of 1e10)
[
0,
# Add directives to the inversion
# ratio to use for each phys prop. smoothness in each direction;
# roughly the ratio of the order of magnitude of each phys. prop.
alpha0_ratio = np.r_[
np.zeros(len(reg.objfcts[0].objfcts)),
1e-4 * np.ones(len(reg.objfcts[1].objfcts)),
100.0 * 1e-4 * np.ones(len(reg.objfcts[2].objfcts)),
]
Alphas = directives.AlphasSmoothEstimate_ByEig(alpha0_ratio=alpha0_ratio, verbose=True)
# initialize beta and beta/alpha_s schedule
beta = directives.BetaEstimate_ByEig(beta0_ratio=1e-2)
betaIt = directives.PGI_BetaAlphaSchedule(
verbose=True, coolingFactor=2.0, tolerance=0.2, progress=0.2,
)
# geophy. and petro. target misfits
targets = directives.MultiTargetMisfits(verbose=True,)
# add learned mref in smooth once stable
MrefInSmooth = directives.PGI_AddMrefInSmooth(wait_till_stable=True, verbose=True,)
# update the parameters in smallness (L2-approx of PGI)
update_smallness = directives.PGI_UpdateParameters(
update_gmm=False # keep GMM model fixed
)
# pre-conditioner
update_Jacobi = directives.UpdatePreconditioner()
# iteratively balance the scaling of the data misfits
scaling_init = directives.ScalingMultipleDataMisfits_ByEig(chi0_ratio=[1.0, 100.0])
scale_schedule = directives.JointScalingSchedule(verbose=True)
# Create inverse problem
# Optimization
# set lower and upper bounds
verbose=True,
n_pw_iter=10)
scaling_schedule = directives.JointScalingSchedule(verbose=True)
alpha0_ratio = np.r_[np.zeros(len(reg_simple.objfcts[0].objfcts)),
100.0 * np.ones(len(reg_simple.objfcts[1].objfcts)),
1.0 * np.ones(len(reg_simple.objfcts[2].objfcts)), ]
alphas = directives.AlphasSmoothEstimate_ByEig(alpha0_ratio=alpha0_ratio,
n_pw_iter=10,
verbose=True)
beta = directives.BetaEstimate_ByEig(beta0_ratio=1e-5, n_pw_iter=10)
betaIt = directives.PGI_BetaAlphaSchedule(
verbose=True,
coolingFactor=2.0,
progress=0.2,
)
targets = directives.MultiTargetMisfits(verbose=True)
petrodir = directives.PGI_UpdateParameters(update_gmm=False)
# Setup Inversion
inv = inversion.BaseInversion(
invProb,
directiveList=[
alphas, scales, beta, petrodir, targets, betaIt, scaling_schedule
],
)
mcluster_map = inv.run(minit)
# Inversion with no nonlinear mapping
reg_simple_no_map = utils.make_SimplePGI_regularization(
mesh=mesh,
