Naming for NuclearNorm - Githubissues

adler-j commented 6 years ago

Basically "NuclearNorm" as currently written has some naming issues:

See doc: https://odlgroup.github.io/odl/generated/odl.solvers.functional.default_functionals.NuclearNorm.html

First, the current definition is actually something like "pointwise nuclear norm and global summation" or something like that. Especially with the new and sexy tensors, we should fix this.
Further, the code is more global than it markets itself. We could go for the more correct name: https://en.wikipedia.org/wiki/Schatten_norm

kohr-h commented 6 years ago

Right, NuclearNorm seems to be associated with inner exponent 1, Schatten norm would be the right generalization. As for the integrated part, I don't know. Couldn't that be called Group Schatten norm, like for the standard p-norms?

mehrhardt commented 6 years ago

I think I wouldn't worry about the integration. This is similar to other norms e.g. L2 where it is clear from the space what exactly is meant by it.

Regarding, Nuclear v Schatten, here I agree that Schatten norm is the better terminology.

odlgroup / odl

Naming for NuclearNorm #1262