domfournier
domfournier
- Overhaul of the Base, Simple and Tikhonov regularizations - Deal with issue #1037
Going through the latest update I noticed a major slowdown of the inversion. I came to realize that `simulation._Jmatrix` gets deleted during the CG iterations every time `self.model` gets updated...
Currently the behaviour between Tikhonov and Simple is different with respect to the cell_weights. - Tikhonov always multiply by cell_volumes - Simple only multiply if cell_weights are None We need...
Currently, the Sparse regularization assumes cell_volumes included in cell_weights. Conflicts with implementation of #1004
The import error seems to occur after updating the simpeg requirements. - Likely caused by mixing channels `conda` and `conda-forge` in the environment Workaround needed.
https://github.com/simpeg/simpeg/blob/d20eefa6b2238210d1c6c14f998c53bb8ea009a4/SimPEG/potential_fields/gravity/simulation.py#L52 @lheagy Why did we switch to a for-loop here?
Installing discretize from pip returns this error on windows: 
New implementation for MVI sparse inversion to keep the problem (mostly) linear. - Implementation of `VectorAmplitude` regularization $\phi_s = \|| \mathbf{W} \mathbf{P} \mathbf{a}(m - m_{ref}) \||_2^2$ - Implementation of `Group`...
#### Summary Refactoring of the UpdateIRLS directive used to control the `Sparse` regularization. - Attribute name changes to snake_case format - Remove unused attributes and change to instance attributes. -...
**This PR is for discussion and planning only** Proposed restructuring of the directives under different modules. The goal is to reduce code duplication across classes and overall improve code quality....