SimPEG 0.21.0 Release Notes#

April 8th, 2024

Updates#

New features#

Gravity simulation using Choclo#

Now we can use a faster and more memory efficient implementation of the gravity simulation SimPEG.potential_fields.gravity.Simulation3DIntegral, making use of Choclo and Numba. To make use of this functionality you will need to install ``choclo` <https://www.fatiando.org/choclo/latest/install.html>`__ in addition to SimPEG.

See simpeg/simpeg#1285.

Use Dask with MetaSimulation#

A new SimPEG.meta.DaskMetaSimulation class has been added that allows to use Dask with SimPEG.meta.MetaSimulations.

See simpeg/simpeg#1199.

Rotated Gradients#

Added a new SimPEG.regularization.SmoothnessFullGradient regularization class that allows to regularize first order smoothness along any arbitrary direction, enabling anisotropic weighting. This regularization also works for a SimplexMesh.

See simpeg/simpeg#1167.

Logistic Sigmoid Map#

New SimPEG.map.LogisticSigmoidMap mapping class that computes the logistic sigmoid of the model parameters. This is an alternative method to incorporate upper and lower bounds on model parameters.

See simpeg/simpeg#1352.

Create Jacobian matrix in NSEM and FDEM simulations#

The frequency domain electromagnetic simulations (including natural source) now support creating and storing the Jacobian matrix. You can access it by using the getJ method.

See simpeg/simpeg#1276.

Documentation#

This new release includes major improvements in documentation pages: more detailed docstrings of classes and methods, the addition of directive classes to the API reference, improvements to the contributing guide, among corrections and fixes.

Breaking changes#

Removal of deprecated bits#

Several deprecated bits of code has been removed in this release. From old classes, methods and properties that were marked for deprecation a few releases back. These removals simplify the SimPEG API and cleans up the codebase.

Remove factor of half in data misfits and regularizations#

Simplify the definition of data misfit and regularization terms by removing the leading factor of one half from these functions. This change makes it easier to interpret the resulting values of these objective functions, while avoiding confusions with their definition.

See simpeg/simpeg#1326.

Bugfixes#

A few bugs have been fixed:

Contributors#

This is a combination of contributors and reviewers who’ve made contributions towards this release (in no particular order).

We would like to highlight the contributions made by new contributors:

Pull Requests#