Alec Hammond

Thanks @ianwilliamson for popping in! (I thought about tagging you earlier, but didn't know how busy you'd be... glad you stumbled you're way across this 😄 ) @mfschubert it might...

(cc @chiaweihsu, @wsshin)

Thanks for the quick reply, @wsshin! > Therefore, I would rather focus on developing a new geometry package Have you already started developing a new package (or is this more...

@stevengj I'm happy to take this on. I've been doing some reading. It _seems_ that Julia's union splitting has [come a long way over the years](https://j-fu.github.io/marginalia/julia/unionize/), such that the compiler...

Thanks for chiming in! > The issue is when we leave Python to do the main compute, for instance the linear solve (Coulomb, heat equation) or eigensolve (mode solving). You...

> For a specific problem we can derive our own adjoint and implement our own PyTorch or JAX primitive for embedding into a backpropagation pipeline. I may be misunderstanding you....

> I'm wondering if we can instanciate this as the AD of a "sparse eigensolver" operator that would work for all sorts of wave problems, instead of an E&M, Schrodinger,...

I agree, @mrbaozi, dealing with the _parameterization_ is tricky. But luckily there are some established solutions here. If you are interested in shape optimization, as opposed to TO, then you...

@HelgeGehring wow, nice work! (I'm a big Julia user, and can't wait to leverage its features in this context!) I think the best place to start is to identify a...

This is great! Thanks for the thorough reply. I'm happy to test this using native GPU arrays. I'm also interested in enabling basic RMA patterns (and can contribute PRs etc)....