Sparse Inversion with Iteratively Re-Weighted Least-Squares#

Least-squares inversion produces smooth models which may not be an accurate representation of the true model. Here we demonstrate the basics of inverting for sparse and/or blocky models. Here, we used the iteratively reweighted least-squares approach. For this tutorial, we focus on the following:

Defining the forward problem

Defining the inverse problem (data misfit, regularization, optimization)

Defining the paramters for the IRLS algorithm

Specifying directives for the inversion

Recovering a set of model parameters which explains the observations

import numpy as np
import matplotlib.pyplot as plt

from discretize import TensorMesh

from simpeg import (
    simulation,
    maps,
    data_misfit,
    directives,
    optimization,
    regularization,
    inverse_problem,
    inversion,
)

# sphinx_gallery_thumbnail_number = 3

Defining the Model and Mapping#

Here we generate a synthetic model and a mappig which goes from the model space to the row space of our linear operator.

nParam = 100  # Number of model paramters

# A 1D mesh is used to define the row-space of the linear operator.
mesh = TensorMesh([nParam])

# Creating the true model
true_model = np.zeros(mesh.nC)
true_model[mesh.cell_centers_x > 0.3] = 1.0
true_model[mesh.cell_centers_x > 0.45] = -0.5
true_model[mesh.cell_centers_x > 0.6] = 0

# Mapping from the model space to the row space of the linear operator
model_map = maps.IdentityMap(mesh)

# Plotting the true model
fig = plt.figure(figsize=(8, 5))
ax = fig.add_subplot(111)
ax.plot(mesh.cell_centers_x, true_model, "b-")
ax.set_ylim([-2, 2])

(-2.0, 2.0)

Defining the Linear Operator#

Here we define the linear operator with dimensions (nData, nParam). In practive, you may have a problem-specific linear operator which you would like to construct or load here.

# Number of data observations (rows)
nData = 20

# Create the linear operator for the tutorial. The columns of the linear operator
# represents a set of decaying and oscillating functions.
jk = np.linspace(1.0, 60.0, nData)
p = -0.25
q = 0.25


def g(k):
    return np.exp(p * jk[k] * mesh.cell_centers_x) * np.cos(
        np.pi * q * jk[k] * mesh.cell_centers_x
    )


G = np.empty((nData, nParam))

for i in range(nData):
    G[i, :] = g(i)

# Plot the columns of G
fig = plt.figure(figsize=(8, 5))
ax = fig.add_subplot(111)
for i in range(G.shape[0]):
    ax.plot(G[i, :])

ax.set_title("Columns of matrix G")

Text(0.5, 1.0, 'Columns of matrix G')

Defining the Simulation#

The simulation defines the relationship between the model parameters and predicted data.

sim = simulation.LinearSimulation(mesh, G=G, model_map=model_map)

Predict Synthetic Data#

Here, we use the true model to create synthetic data which we will subsequently invert.

# Standard deviation of Gaussian noise being added
std = 0.02
np.random.seed(1)

# Create a SimPEG data object
data_obj = sim.make_synthetic_data(true_model, noise_floor=std, add_noise=True)

Define the Inverse Problem#

The inverse problem is defined by 3 things:

Data Misfit: a measure of how well our recovered model explains the field data

Regularization: constraints placed on the recovered model and a priori information

Optimization: the numerical approach used to solve the inverse problem

# Define the data misfit. Here the data misfit is the L2 norm of the weighted
# residual between the observed data and the data predicted for a given model.
# Within the data misfit, the residual between predicted and observed data are
# normalized by the data's standard deviation.
dmis = data_misfit.L2DataMisfit(simulation=sim, data=data_obj)

# Define the regularization (model objective function). Here, 'p' defines the
# the norm of the smallness term and 'q' defines the norm of the smoothness
# term.
reg = regularization.Sparse(mesh, mapping=model_map)
reg.reference_model = np.zeros(nParam)
p = 0.0
q = 0.0
reg.norms = [p, q]

# Define how the optimization problem is solved.
opt = optimization.ProjectedGNCG(
    maxIter=100, lower=-2.0, upper=2.0, maxIterLS=20, maxIterCG=30, tolCG=1e-4
)

# Here we define the inverse problem that is to be solved
inv_prob = inverse_problem.BaseInvProblem(dmis, reg, opt)

Define Inversion Directives#

Here we define any directiveas that are carried out during the inversion. This includes the cooling schedule for the trade-off parameter (beta), stopping criteria for the inversion and saving inversion results at each iteration.

# Add sensitivity weights but don't update at each beta
sensitivity_weights = directives.UpdateSensitivityWeights(every_iteration=False)

# Reach target misfit for L2 solution, then use IRLS until model stops changing.
IRLS = directives.Update_IRLS(max_irls_iterations=40, minGNiter=1, f_min_change=1e-4)

# Defining a starting value for the trade-off parameter (beta) between the data
# misfit and the regularization.
starting_beta = directives.BetaEstimate_ByEig(beta0_ratio=1e0)

# Update the preconditionner
update_Jacobi = directives.UpdatePreconditioner()

# Save output at each iteration
saveDict = directives.SaveOutputEveryIteration(save_txt=False)

# Define the directives as a list
directives_list = [sensitivity_weights, IRLS, starting_beta, update_Jacobi, saveDict]

Setting a Starting Model and Running the Inversion#

To define the inversion object, we need to define the inversion problem and the set of directives. We can then run the inversion.

# Here we combine the inverse problem and the set of directives
inv = inversion.BaseInversion(inv_prob, directives_list)

# Starting model
starting_model = 1e-4 * np.ones(nParam)

# Run inversion
recovered_model = inv.run(starting_model)

Running inversion with SimPEG v0.22.1

                    simpeg.InvProblem is setting bfgsH0 to the inverse of the eval2Deriv.
                    ***Done using the default solver Pardiso and no solver_opts.***

model has any nan: 0
=============================== Projected GNCG ===============================
  #     beta     phi_d     phi_m       f      |proj(x-g)-x|  LS    Comment
-----------------------------------------------------------------------------
x0 has any nan: 0
1.70e+06  3.72e+03  1.01e-09  3.72e+03    1.99e+01      0
8.49e+05  1.87e+03  3.86e-04  2.20e+03    1.93e+01      0
4.25e+05  1.28e+03  8.90e-04  1.66e+03    1.88e+01      0   Skip BFGS
2.12e+05  7.48e+02  1.78e-03  1.13e+03    1.80e+01      0   Skip BFGS
1.06e+05  3.74e+02  3.02e-03  6.94e+02    1.65e+01      0   Skip BFGS
5.31e+04  1.63e+02  4.39e-03  3.96e+02    1.42e+01      0   Skip BFGS
2.65e+04  6.65e+01  5.64e-03  2.16e+02    1.27e+01      0   Skip BFGS
1.33e+04  2.90e+01  6.61e-03  1.17e+02    1.01e+01      0   Skip BFGS
Reached starting chifact with l2-norm regularization: Start IRLS steps...
irls_threshold 1.2669107150802716
6.63e+03  1.60e+01  1.00e-02  8.25e+01    4.05e+00      0   Skip BFGS
1.12e+04  1.46e+01  1.15e-02  1.43e+02    1.69e+01      0
8.79e+03  2.43e+01  1.15e-02  1.25e+02    2.91e+00      0
6.86e+03  2.46e+01  1.23e-02  1.09e+02    3.25e+00      0   Skip BFGS
5.45e+03  2.38e+01  1.28e-02  9.35e+01    4.40e+00      0   Skip BFGS
4.47e+03  2.24e+01  1.29e-02  8.01e+01    4.58e+00      0   Skip BFGS
4.47e+03  2.07e+01  1.26e-02  7.69e+01    6.10e+00      0
4.47e+03  2.07e+01  1.17e-02  7.30e+01    6.39e+00      0
4.47e+03  2.06e+01  1.07e-02  6.84e+01    6.50e+00      0
4.47e+03  2.02e+01  9.64e-03  6.33e+01    6.84e+00      0
4.47e+03  1.97e+01  8.56e-03  5.79e+01    7.31e+00      0
4.47e+03  1.89e+01  7.41e-03  5.21e+01    7.79e+00      0
6.97e+03  1.79e+01  6.47e-03  6.30e+01    1.50e+01      0
6.97e+03  2.03e+01  5.31e-03  5.73e+01    1.04e+01      0
6.97e+03  1.98e+01  4.57e-03  5.17e+01    1.03e+01      0   Skip BFGS
6.97e+03  1.90e+01  4.11e-03  4.76e+01    1.13e+01      0
6.97e+03  1.84e+01  3.75e-03  4.45e+01    1.19e+01      0
6.97e+03  1.87e+01  3.20e-03  4.10e+01    1.77e+01      0
1.09e+04  1.77e+01  2.71e-03  4.72e+01    1.70e+01      0
1.09e+04  1.87e+01  2.17e-03  4.24e+01    1.35e+01      0
1.09e+04  1.82e+01  1.80e-03  3.79e+01    1.26e+01      0
1.71e+04  1.77e+01  1.50e-03  4.34e+01    1.87e+01      0
1.71e+04  1.85e+01  1.19e-03  3.88e+01    1.41e+01      0
1.71e+04  1.85e+01  1.02e-03  3.59e+01    1.64e+01      1   Skip BFGS
1.71e+04  1.89e+01  8.52e-04  3.34e+01    1.35e+01      0
1.71e+04  1.92e+01  7.23e-04  3.15e+01    1.32e+01      0   Skip BFGS
1.71e+04  1.95e+01  6.14e-04  3.00e+01    1.37e+01      0
1.71e+04  1.98e+01  5.23e-04  2.87e+01    1.39e+01      0
1.71e+04  2.01e+01  4.46e-04  2.77e+01    1.41e+01      0
1.71e+04  2.03e+01  3.80e-04  2.68e+01    1.42e+01      0
1.71e+04  2.05e+01  3.17e-04  2.59e+01    1.40e+01      0
1.71e+04  2.05e+01  2.66e-04  2.51e+01    1.40e+01      0
1.71e+04  2.06e+01  2.26e-04  2.44e+01    1.38e+01      0
1.71e+04  2.06e+01  1.89e-04  2.38e+01    1.37e+01      0
1.71e+04  2.06e+01  1.70e-04  2.35e+01    1.68e+01      2
1.71e+04  2.07e+01  1.53e-04  2.33e+01    1.76e+01      2   Skip BFGS
1.71e+04  2.11e+01  1.17e-04  2.31e+01    1.72e+01      0
1.71e+04  2.11e+01  9.65e-05  2.27e+01    1.36e+01      0
1.71e+04  2.11e+01  8.97e-05  2.26e+01    1.64e+01      8   Skip BFGS
1.71e+04  2.11e+01  8.29e-05  2.25e+01    1.78e+01      3
Reach maximum number of IRLS cycles: 40
------------------------- STOP! -------------------------
: |fc-fOld| = 0.0000e+00 <= tolF*(1+|f0|) = 3.7168e+02
: |xc-x_last| = 4.0540e-02 <= tolX*(1+|x0|) = 1.0010e-01
: |proj(x-g)-x|    = 1.7789e+01 <= tolG          = 1.0000e-01
: |proj(x-g)-x|    = 1.7789e+01 <= 1e3*eps       = 1.0000e-02
: maxIter   =     100    <= iter          =     48
------------------------- DONE! -------------------------

Plotting Results#

fig, ax = plt.subplots(1, 2, figsize=(12 * 1.2, 4 * 1.2))

# True versus recovered model
ax[0].plot(mesh.cell_centers_x, true_model, "k-")
ax[0].plot(mesh.cell_centers_x, inv_prob.l2model, "b-")
ax[0].plot(mesh.cell_centers_x, recovered_model, "r-")
ax[0].legend(("True Model", "Recovered L2 Model", "Recovered Sparse Model"))
ax[0].set_ylim([-2, 2])

# Observed versus predicted data
ax[1].plot(data_obj.dobs, "k-")
ax[1].plot(inv_prob.dpred, "ko")
ax[1].legend(("Observed Data", "Predicted Data"))

# Plot convergence
fig = plt.figure(figsize=(9, 5))
ax = fig.add_axes([0.2, 0.1, 0.7, 0.85])
ax.plot(saveDict.phi_d, "k", lw=2)

twin = ax.twinx()
twin.plot(saveDict.phi_m, "k--", lw=2)
ax.plot(np.r_[IRLS.iterStart, IRLS.iterStart], np.r_[0, np.max(saveDict.phi_d)], "k:")
ax.text(
    IRLS.iterStart,
    0.0,
    "IRLS Start",
    va="bottom",
    ha="center",
    rotation="vertical",
    size=12,
    bbox={"facecolor": "white"},
)

ax.set_ylabel(r"$\phi_d$", size=16, rotation=0)
ax.set_xlabel("Iterations", size=14)
twin.set_ylabel(r"$\phi_m$", size=16, rotation=0)

Text(865.1527777777777, 0.5, '$\\phi_m$')

Total running time of the script: (0 minutes 27.241 seconds)

Estimated memory usage: 9 MB

Download Jupyter notebook: plot_inv_2_inversion_irls.ipynb

Download Python source code: plot_inv_2_inversion_irls.py

Gallery generated by Sphinx-Gallery