[Soplex] performance, warm start, and early termination

[Zohar Levi]

Okay, thanks.

On 17/8/22 5:45, Ambros Gleixner wrote:
> Hi Zohar,
>
> 1. SoPlex may take longer than the commercial solvers; it may even be 
> that they use interior point or concurrent instead of pure simplex.  
> But unsimplify should not take that long, this must be a performance 
> bug. You can file a bug report on github.com/scipopt/soplex with log 
> and precise info how to reproduce it.
>
> 2. You can provide a so called basis file on the command line, but not 
> a primal solution.
>
> 3. No, you will not be able to achieve what you describe in the 
> stackexchange issue with soplex, since intermediate solutions are not 
> guaranteed to be feasible.  SoPlex applies what is called a composite 
> simplex, switching between primal and dual, plus adds intermediate 
> perturbations for numerical stability.
>
> Hope that helps,
> Ambros
>
>
> Am 12.08.22 um 16:40 schrieb Zohar Levi:
>>
>> This is my objective:
>>
>> https://math.stackexchange.com/questions/4510587/find-a-feasible-but-not-optimal-solution-to-a-large-lp 
>>
>>
>> 1. I tried soplex on a small problem:
>>
>> https://drive.google.com/file/d/1RhCqdziUNbrXJ1UyArQtIZevbXV2Shg-/view
>>
>> soplex mymodel.mps
>>
>> It found an optimal solution after 2min, and it spent another 8min to 
>> unsimplify it.
>>
>> Mosek and gurobi needed 2sec.
>>
>> What am I doing wrong?
>>
>> 2. How do I specify a warm-start (an initial feasible guess)? In the 
>> main page, there's a mention of hot-start, but I couldn't find it in 
>> the parameters on anywhere else. Something lost in translation?
>>
>> 3. If I start from a feasible solution and stop the solver after X 
>> iterations, will I get a better feasible solution?
>>
>>
>> _______________________________________________
>> SoPlex mailing list
>> SoPlex at zib.de
>> https://listserv.zib.de/mailman/listinfo/soplex