Remarks as a user :)

[aTrotier]

Hi,
I will use that topic to make some remarks of stuff that feels weird to me.

coherence in dimensions

Most of the time the dimensions of an array are (sx,sy,sz,echos,channel)

Because I'm a BART user It is already weird for me ^^, but something less intuitive is that the dimensions for the sensitivity maps has to be (sx,sy,sz,channel) in the reconstruction function

I suppose one of the reasons is : sens[:,:,:,1,:] remove the echo dimension

[aTrotier]

The compressed sensing example (with TV) in the doc is not working in 3D.
The proxTV operator only works for 2D, no ?

[tknopp]

Probably. Jakob already generalized quite a bit of code in RegularizedLeastSquares to ND.