I attended the AI WILD workshop on inverse problems and imaging. I heard very good talks there, I think the one I liked the most was from Pierre Weiss, presenting his work with Hai on Analytical solutions for CNN inverse problem solvers! I like his honest views, and the fact he uses plain language. Of the course the talk of Julian was also nice but I think I heard it 5 times by now, although this time it had a bit of RAM. The contributed presentation of Alessio Spagnoletti on latino-pro was also cool
On my side, I presented for the firs time what I’ve been cooking for 6 months now MLEM as a Bregman mirror descent. I met Alix Chazottes who told me she was working on similar stuff, her work on variable metric Bremgan methods ended up being quite different. I also presented my work to Antonin Clerc and Nicolas Papadakis who were actually working on exactly the same thing ! It seems they were a few months behind though, they thought the reparameterized objective was convex: been there done that ! As of late September 2026 my pre-print on this is still not out, while theirs was already out in February 🥲 To be fair, the work has evolved into something quite different in the mean time.
