Exactly; it's about generating (much faster) heuristics for "guessing" the outcome given certain parameters, instead of running the real simulation every time.

Exactly; it's about generating (much faster) heuristics for "guessing" the outcome given certain parameters, instead of running the real simulation every time. Useful, for sure, but the real thing should be used to verify once you've found something interesting.