[Coin-ipopt] Re: Bug report

[Zhiwen Chong]

On 5-Apr-06, at 10:23 AM, Andreas Waechter wrote:
> Setting a slightly positive bound on that variable might help.   
> Note, by default Ipopt relaxes the user-given bounds slightly (by  
> about 1e-8) before it solves the problem, so that if you put 0 as  
> bound, it is actually slightly negative.  Setting the option  
> bound_relax_factor to 0 makes Ipopt not change the original bounds.

That's useful to know.

> That is strange.  Did you download the MC19 code from Harwell and  
> put it into Extern/HSL/mc19ad.f ?  If you didn't, the option is  
> ignored, otherwise it's strange that it is ignored (well, it is  
> currently not an option one can set through AMPL, but that will  
> change in the new release. However, providing the option in  
> PARAMS.DAT should have worked).  If you download it now, you need  
> to run configure again and recompile for Ipopt to include this  
> scaling method.

Yes, I had mc19ad.f compiled into my Ipopt 3.0.1.

Well, I just downloaded and compiled the latest Ipopt (3.10) and put  
all my solver parameters in ipopt.opt, including the  
linear_system_scaling option. Now, it seems that this version of  
Ipopt doesn't print out a list of the currently active options when  
it runs, but I believe it *is* reading the ipopt.opt file because  
when when I change the print_level, the effect is immediately visible.

So I think the linear_system_scaling option is *probably* active.

>>> I hope this helps,
>> Yup it does, thanks!
> ...?  So, you were able to solve the problem?

Yup, based on your advice, I reformulated the problem with slightly  
positive bounds (plus some other relaxations + scaling) and it solves  
correctly now. It was a bad formulation to start with.

Thanks!