There are problems that I’m interested in such as inference of PDE/DDE that lack of gradient is a norm for existing solvers. I don’t know if HINTS is the answer but there’s certainly a market for non-gradient-based samplers.
1 Like