NanoComp / meep

free finite-difference time-domain (FDTD) software for electromagnetic simulations
GNU General Public License v2.0
1.19k stars 610 forks source link

Better document the procedure for determining TO lengthscale constraint hyperparameters #2517

Open smartalecH opened 1 year ago

smartalecH commented 1 year ago

As explored throughout #2497, it's rather difficult to determine the proper hyperparameters for the geometrical lengthscale constraints we use (as described in our paper).

It would be good to document what kind of procedure one could follow to determine the kinds of thresholds one needs in order to properly implement the constraint. For example, you can run a simple mini optimization for a series of geometric sizes, lengthscale constraints, etc like we do here and analyze an appropriate threshold.

In theory, would could build a tool to automate this process. But for now we should at least describe this in the documentation.

(cc @oskooi, @mfschubert )