[Ipopt] Run optimizer with 1 preferred starting point

Stefan Vigerske stefan at math.hu-berlin.de
Sat Nov 24 12:04:51 EST 2018


once you call Ipopt, it will try to do one optimization with the 
starting point that you provide via your implementation of the 
TNLP::get_starting_point() function.
If you want Ipopt to run more often, then you need to call the 
IpoptApplication::OptimizeTNLP() function more often and then provide 
the corresponding starting point in TNLP::get_starting_point(). To speed 
things up a bit, you can actually call 
IpoptApplication::ReOptimizeTNLP() instead of 
IpoptApplication::OptimizeTNLP() for all by the first call.


On 11/18/18 5:07 PM, R zu wrote:
> Hi.
> - I am using ipopt as part of an iterative algorithm.
> - I want Ipopt to optimize an objective function with many variables.
> - The old solution from last iteration (x0) of the overall algorithm is a
> good starting point for the new solution.
> - I have already wrote a function to generate random starting point.
> Can I ask Ipopt to just optimize for once with 1 starting point that I
> provide (x0)?
> Can I ask Ipopt to try to start from x0 first and then try other starting
> point?
> Thank you.
> _______________________________________________
> Ipopt mailing list
> Ipopt at list.coin-or.org
> https://list.coin-or.org/mailman/listinfo/ipopt

More information about the Ipopt mailing list