uqfoundation / mystic

constrained nonlinear optimization for scientific machine learning, UQ, and AI
http://mystic.rtfd.io
Other
467 stars 50 forks source link

add high-level ability to configure and run staged optimizations #190

Open mmckerns opened 1 year ago

mmckerns commented 1 year ago

Currently, if one wanted to perform a staged optimization it's a bit of a manual process. One can, for example, start with a four-parameter optimization, which runs until termination... then, if the class interface was used, one can change the termination conditions, constraints, or other parts of the optimizer state, and restart the optimization, now using ten parameters. This may require the latter six parameters to be initially pinned with the bounds, a constraint, or a penalty, as opposed to directly expanding the optimizer from four to ten parameters.

It's be beneficial to create a function or class that contains a plan for easily automating a staged optimization.