SimPEG 0.15.2 Release Notes#
November 12th, 2021
This patch release contains some improvements and bug fixes for the code base.
Updates#
Regularization Cell Weights#
Cell weights for all regularizations now are unified in their treatment. We do not assume that the cell weights have been multiplied by volumes ahead of time.
Along with this new assumption, we have updated the sensitivity weighting directive to
normalize the sensitivity of J.T * J
by volume. This allows a more intuitive definition of
new weighting functions. We do expect this to slightly change examples that were using
the Simple
and Sparse
regularization classes. If you were using set weighting
parameters (i.e. not starting with a beta estimator directive), you will likely need to
choose new weighting parameters.
PGI#
There are several incremental improvements to the petrophysical guided inversion (PGI) regularizations and directives, along with a bug fix to support the latest version of scikit-learn.
Others#
We have added support for reading in falcon formatted UBC-GIF styled data gravity gradient data. We have also added a bug fix for reading in UBC-GIF IP data.
We have fixed a bug with TreeMesh
being used for the Simple
regularization when
not specifically specifying a active cells list.
There is now a forward_only
option for the frequency domain methods to discard any
stored matrix factorizations upon completion.
Contributors#
This is a combination of contributors and reviewers who’ve made contributions towards this release (in no particular order).
Pull requests#
#1038: Issue-1037
#1044: Add forward only option to 3D FDEM
#1045: No active cells bug in regularization mesh
#1046: Fix bug in IO for DCIP3D data
#1050: Fix for scikit-learn 1.0 (back compatible with scikit-learn 0.24.2)
#1051: add Falcon Guv component to gg3d io_utils.
#1053: PGI Regularization Refactor