2025-03-26 Gokey/Mitchell/Wang/Wagner Check-in meeting notes

(11 mins)

JM – If we want to do broad searches that quickly discard bad candidates, how would we want to do that?

• TG – Would need clearer definitions of what you’re looking for. The user will need to specify these. Ex could just use elements as primitives but even then you’ll have a huge number of options.

• JM – Could make a really fast objective to prune things

• TG – Could make tiers that don’t have to do fits, kinda like single point properties. Resetting to sage parameters after splitting could offer a speedup too, but is kinda a science question. Would turn on step_limit=0 and reset.

…

TG – Can only do step_limit=0 if you have hessians

• LW – We don’t have hessians now but can generate them easily

(detailed technical stuff, see recording)