Dan Stahlke

I did track down a description of this Schur trick for logdet: section 6.2.3 of https://docs.mosek.com/modeling-cookbook/sdo.html One could also consider [I U ; U' X] in :SDP, which is feasible...

I'm interested in relative entropy. That Fawzi paper and module looks promising. It's only an approximation, but it looks like they are getting to within 1e-5 or 1e-6 which is...

Thank you for that detailed description. I'm coming to learn that one of the big strengths of Julia is its interoperability and well designed APIs and ecosystem. I'll do some...

That's great news. I was prepared to reproduce it from the paper, but of course looking at code makes it easier and more likely to be correct. I have so...

I've started porting matrix_geo_mean_hypo_cone because all the other functions depend on this one. Question though: is it possible for an atom to return a matrix variable rather than a scalar?...

Thanks. This helps me get started. I've implemented a couple of the atoms, but haven't tested much. It would be nice if the cones can be atoms because it leads...

Wow, congrats on the thesis submission! Bet it feels good to be done with that. I've been saying "then I'll have some free time" for a long time now, and...

So, I experimented a bit with making GeomMeanHypoCone a constraint rather than a function: https://github.com/dstahlke/Convex.jl/tree/logm-constraint ... and I'm stuck right in the middle of indecisiveness about whether this is better....

Just a quick update: I have this mostly complete, but keep getting sidetracked with higher priority tasks. In case anybody coming across this thread needs this functionality: let me know...

For now the Semaphore workaround works for me. Perhaps I will one day find time to find a fix and send you a pull request.