The PGRA model (= Gauss-Newton optimisation of W, Z and R) was refactored. It is now organised the same way as the PGVA model (= Closed-form updates of W, Z from V).
There is still some work to do:
merge pgva_scale_pg and pgra_scale_pg into a unique, modular, function
maybe simplify lower bound tracking
update/write "register" functions (= apply a learnt model to a (or several) new images)
The PGRA model (= Gauss-Newton optimisation of W, Z and R) was refactored. It is now organised the same way as the PGVA model (= Closed-form updates of W, Z from V).
There is still some work to do: