Open ccoffrin opened 5 years ago
My intention is models that are more complex than quadratic but only use the following terms in arithmetic expressions -
,+
,*
,^
(and maybe \
,/
). Discrete variables will be considered as well.
can I suggest being a little more specific clearer than "polynomial models", since we have PolyJuMP for a very very different type of "polynomial optimization". if you are allowing general powers of variables (ie x^a
with a
possibly non-integer) then "signomial" would be more accurate. otherwise perhaps "multilinear".
multilinear might be most accurate, but I'll leave it to @kaarthiksundar to comment on the specifics. The goal is to have tests for solvers that use the NLP interface but also don't support all of MINLP, for example transcendental functions. This includes solvers like Alpine, Baron, and SCIP.
How are you defining polynomial models here?