[Ipopt] Why ipopt needs so many iterations when starting from the optimal solution?

[Gianpaolo Gulletta]

Is this the reason of the problem? 

The problem is that IPOPT either takes many iterations or needs much time when I provide a starting point that is close to the optimal point. Why? It sounds strange to me, IPOPT should perform better when it starts close to its optimal solution.

 

Are there any settings that can solve the problem?

 

Thank you.

 

Gianpaolo Gulletta

 

From: Ipopt User [mailto:ipoptuser at gmail.com] 
Sent: Monday, June 29, 2015 12:52 PM
To: Gianpaolo Gulletta
Subject: Re: [Ipopt] Why ipopt needs so many iterations when starting from the optimal solution?

 

Please send your response to ipopt at list.coin-or.org <mailto:ipopt at list.coin-or.org> 

 

But yes, that is the problem, and there is no easy workaround if you do not calculate the Hessian  yourself.

 

On Mon, Jun 29, 2015 at 3:48 AM, Gianpaolo Gulletta <gianpaologulletta at gmail.com <mailto:gianpaologulletta at gmail.com> > wrote:

Is this the reason of the problem? 

The problem is that IPOPT either takes many iterations or needs much time when I provide a starting point that is close to the optimal point. Why? It sounds strange to me, IPOPT should perform better when it starts close to its optimal solution.

 

Are there any settings that can solve the problem?

 

Thank you.

 

Gianpaolo

 

From: Ipopt User [mailto:ipoptuser at gmail.com <mailto:ipoptuser at gmail.com> ] 
Sent: 29 June 2015 03:01 AM
To: Gianpaolo Gulletta
Cc: ipopt at list.coin-or.org <mailto:ipopt at list.coin-or.org> 
Subject: Re: [Ipopt] Why ipopt needs so many iterations when starting from the optimal solution?

 

You use 'hessian_approximation','limited-memory', which means at the start it has no idea about the Hessian. Moreover, you probably do not set the values of the dual variables. Only if IPOPT knows the Hessian and both primal and dual values, it can take the most sensible steps towards the optimal solution.

 

On Sat, Jun 27, 2015 at 12:13 PM, Gianpaolo Gulletta <gianpaologulletta at gmail.com <mailto:gianpaologulletta at gmail.com> > wrote:

Hi everybody,

 

I realized that the performance of the algorithm Ipopt decreases when it starts either from the optimal solution or from very closely.

I saw that the problem has already been addressed here <https://www.or-exchange.org/questions/1268/why-ipopt-needs-so-many-iterations-when-starting-from-the-optimal-solution>  but I am still looking for the answer.

 

Can anyone help? 

 

Thank you.

 

Regards,

 

Gianpaolo Gulletta

 


_______________________________________________
Ipopt mailing list
Ipopt at list.coin-or.org <mailto:Ipopt at list.coin-or.org> 
http://list.coin-or.org/mailman/listinfo/ipopt

 

 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.coin-or.org/pipermail/ipopt/attachments/20150629/0b0d3b07/attachment.html>