Closed ChristophWWagner closed 5 years ago
I propose to remove MLUltraSound altogether.
We also need to figure out, if we want to talk about Circulant Matrices, or if we talk about circular convolution matrices, at least in the docu. The same holds for Toeplitz Matrices, which ultimately are just convolution matrices. This way we do not have to talk about the number of levels, but instead the number of levels corresponds to the dimension of the convolution kernels.
So, I would describe the matrices as the transforms, that carry out convolutions with L-dimensional convolution kernels, so it is implicitly assumed, that the data is a reshaping of some L-dimensional object.
MLToeplitz can now deal with arbitrary shapes of each level. It is also implemented in such a way that it allows to be used as a standard Toeplitz class. Moreover, it now hosts specialized column and row normalization routines.
As we have a separate PR for documentation improvement the documentation rework will be done in that PR, so the new feature improvement will roll out to alpha stage and doc hitting the pipeline before leaving the 0.2pre stage
As the multi-level extensions of some classes are by look and feel match exactly their single level counterpart it makes sense to just merge them.
In particular:
What needs to be adapted