(j3.2006) Liaison to IFIP WG 2.5

Aleksandar Donev donev1
Wed Aug 22 16:29:16 EDT 2007


On Wednesday 22 August 2007 12:32, Van Snyder wrote:
> The poster child for this is global minimization.
Yes, but the so-called globally convergent interval methods (branch and bound 
with intervals) have not been shown to actually work in practice other than 
on simple problems in low dimensions. I've tried them myself a few times 
(using the Fortran software implementation) and they were too slow (the 
number of branches may still be exponential so just because you can get the 
answer doesn't mean you will in a reasonable time). If any technology really 
solved the global optimization problem it would be the next Big Thing. I am 
afraid we are very far off still...
Best,
Aleks



More information about the J3 mailing list