search

hdrake / OptimizeClimate

MOVED TO http://github.com/ClimateMARGO/ClimateMARGO.jl
http://github.com/ClimateMARGO/ClimateMARGO.jl
1 stars 0 forks source link

Bug in optimization when cost function concave (p<1) #20

Open hdrake opened 4 years ago

hdrake commented 4 years ago

The Ipopt optimization does not seem to be converging when the cost functions are concave.

I will need to look into the documentation some more: https://coin-or.github.io/Ipopt/OUTPUT.html https://coin-or.github.io/Ipopt/

ron-rivest commented 4 years ago

Hi Henri --

Yes, perhaps there is a bug in Ipopt.

When p<1 the cost functions are not convex, and the global optimum may be on the border of the feasible region. I don't have much familiarity with Ipopt but it should be able to handle this case (at to the extent of giving you a local optimum)...

Let us know how it goes...

Cheers, Ron

On Mon, Jan 27, 2020 at 7:04 AM Henri Drake notifications@github.com wrote:

The Ipopt optimization does not seem to be converging when the cost functions are concave.

I will need to look into the documentation some more: https://coin-or.github.io/Ipopt/OUTPUT.html https://coin-or.github.io/Ipopt/

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/hdrake/OptimizeClimate/issues/20?email_source=notifications&email_token=ABQ765N6STJBTQTB56BGMKTQ73ETHA5CNFSM4KMADMV2YY3PNVWWK3TUL52HS4DFUVEXG43VMWVGG33NNVSW45C7NFSM4II4XDIA, or unsubscribe https://github.com/notifications/unsubscribe-auth/ABQ765KBP6XEDF5ZVX7EJ6DQ73ETHANCNFSM4KMADMVQ .

-- The climate crisis isn't about who's right. It's about who's helping.

hdrake commented 4 years ago

I need to put an example of what happens for p<1