[Coin-discuss] osi, interior

[Matthew Galati]

>
>
>> Hi,
>>
>> Several of the LP solvers in Osi have interior point methods. Can 
>> this be added as an option to solve with interior vs simplex? I guess 
>> the OSI2 design - where model and algorithm are split would fix 
>> this... but, is OSI2 still going to happen? and, if there is no real 
>> timeline for OSI2, can this be added to OSI?
>>
>> Matt
>
>
> It is definitely in the works for OSI2.  There is still not an ETA for 
> OIS2, though.  I'd have to look at what's involved in adding it to the 
> current OSI.  If it's not too hard, I can add to the base class, then 
> I'd pass the solver layer additions to solver layer maintainers.
>
> I'll take a look this week.  Do you have suggestions for must-have 
> features beyond setting a starting point or using the built-in one, 
> invoking the IP solver, and getting back the solution?
>

That should be good enough to start. Also, maybe the ability to do 
crossover (if the solver provides this feature).

Matt