ProxToobox issueshttps://gitlab.gwdg.de/nam/ProxPython/-/issues2021-01-31T16:28:21Zhttps://gitlab.gwdg.de/nam/ProxPython/-/issues/11NoLips TO DO2021-01-31T16:28:21ZRussell LukeNoLips TO DOCurrent structure:
nolipsExperiment.py is the top-level experiment class in this folder.
randomExperiment.py is onle level down and will inherit the class defined by nolipsExperiment.py
I imagine eventually having a JWSTExperim...Current structure:
nolipsExperiment.py is the top-level experiment class in this folder.
randomExperiment.py is onle level down and will inherit the class defined by nolipsExperiment.py
I imagine eventually having a JWSTExperiment.py with, say, the JWST data, and maybe a sensor localization one too.
But we will start with randomly generated quadratics. The only problem with this is that we don't know what the solution is.
# TO DO:
* Algorithm initialization and
* putting in the functions f and phi (see Teboulle notes).
Not sure yet how to do this since these could vary. Initialization could be
done at the individual experiment level. f and phi should probably be coded
at the level of nolipsExperiment.py, which is the top-level experiment class for the NoLips folder.
These are in principle independent of the particular quadratic forms, but we would want these
to be parameters with defaults that the user could change.lausterlausterhttps://gitlab.gwdg.de/nam/ProxPython/-/issues/1simplification of evaluateChange in gretchko2020-06-17T16:07:42ZGijsbert Simon Matthijs Jansensimplification of evaluateChange in gretchkoI think the whole if loop can be avoided if you use something like
np.sum(abs(u-u_new)**2).
This should work for arrays (u, u_new) of arbitrary size and dimension - it will simply sum over all.I think the whole if loop can be avoided if you use something like
np.sum(abs(u-u_new)**2).
This should work for arrays (u, u_new) of arbitrary size and dimension - it will simply sum over all.