 |
     |
|
|
||
|
|
|
|||||||||||||
|
NLBranch Mixed Integer Nonlinear Programming Solver for Metacomputing Platforms. Jean-Pierre Goux, Sven Leyffer and Jorge Nocedal Our goal was to develop a solver able to solve large Mixed-Integer Nonlinear Programming (MINLP) problems on metacomputing platforms. MINLP problems can be modelled in the following form : 
MINLP problems arise in numerous applications areas such as chemical engineering (batch plant design), telecommunication industry (network design) or financial engineering (portfolio revision). MINLP problems are still considered as extremly hard problems since they combine the numerical difficulties of nonlinear programming with the combinatorial aspect of integer programming. Using MW, we developed a parallel branch-and-bound solver running on metacomputing platforms. Unsolved problems from the MINLPLIB collection have been solved and we are investigating ways to solve even larger instances. More information about MW-MINLP :
|
