search

jump-dev / JuMP.jl

Modeling language for Mathematical Optimization (linear, mixed-integer, conic, semidefinite, nonlinear)
http://jump.dev/JuMP.jl/
Other
2.2k stars 392 forks source link

Question: how does JuMP compare with OpenMDAO #2153

Closed mantielero closed 4 years ago

mantielero commented 4 years ago

Do they fill the same niche?

mlubin commented 4 years ago

This is not a bug report or feature request for JuMP, so https://discourse.julialang.org/c/domain/opt would be a better place for this discussion. I'm hesitant to say much because I've never seen OpenMDAO before; I'd rather answer a more specific question about "does JuMP do X?" and let you formulate an opinion on how they fill the niche you care about. JuMP fills a lot of niches beyond nonlinear optimization.

In any case, after a very brief look it looks like OpenMDAO does not implement automatic differentiation as JuMP does; instead, the only options are finite differences or hand-written derivatives:

http://openmdao.org/twodocs/versions/latest/basic_guide/first_mdao.html

We’re using finite difference here for simplicity, but for larger models, finite differencing results in a high computational cost, and can have limited accuracy. It’s much better to use analytic derivatives with your models. You can learn more about that in the Advanced User Guide.

mantielero commented 4 years ago

Thanks closing it. I missed the discourse link. Sorry.