Dear CASToR users,
New features were pushed into the develop branch concerning optimization algorithms:
- a mechanism for optimization algorithms making use of sub steps or sub iterations (such as MLAA or ADMM)
- a mechanism for processing of data before and after optimization for any pre or post processing step necessitating a loop over data elements,
- the possibility to provide additional input in the data space.
Moreover, one penalty and two optimizers were added thanks to these new features:
- the APPGML algorithm from Millardet et al (shifted version of the PPGML algorithm from Nuyts et al),
- the ADMM from Lim et al (maximum likelihood with positivity constraint in the projection space),
- the simple quadratic penalty.
We are working on implementing MLAA optimization algorithm, but some work still need to be done.
The documentation about optimizers and penalties was updated according to these new features. Do not hesitate to contact me for any question !
Kind regards,
Alexandre