doing the integration of OpenMEEG in MNE and comparing solution between MNE linear collocation method and OpenMEEG results seem pretty different when using a 1 layer BEM model.
See https://gist.github.com/agramfort/c9dd2a643a5e02155c23aca915351e79
For a 3 layers model the MAGs and RDMs really good. But for 1 layer (MEG only) it's not great.
I have no intuition why. Although using a finer mesh for the single interface helps (see notebook at the bottom) it's still pretty bad...
@papadop @mclerc if you have any idea why this can happen let me know
also I think we do have any test in CTest/CMake except for 3 layers models. @papadop do you confirm?
doing the integration of OpenMEEG in MNE and comparing solution between MNE linear collocation method and OpenMEEG results seem pretty different when using a 1 layer BEM model.
See https://gist.github.com/agramfort/c9dd2a643a5e02155c23aca915351e79
For a 3 layers model the MAGs and RDMs really good. But for 1 layer (MEG only) it's not great.
I have no intuition why. Although using a finer mesh for the single interface helps (see notebook at the bottom) it's still pretty bad...
@papadop @mclerc if you have any idea why this can happen let me know
also I think we do have any test in CTest/CMake except for 3 layers models. @papadop do you confirm?