Abstract
We present an intelligent branch-and-bound algorithm and report preliminary computational results. The conventional approach to use of a particular branching strategy throughout is replaced by a branching algorithm that uses a neural network to direct the branching. The technique is shown to be significantly faster than conventional methods for certain classes of mixed integer linear programming (MILP) models.
Original language | English |
---|---|
Pages (from-to) | 461-468 |
Number of pages | 8 |
Journal | Computers and Operations Research |
Volume | 19 |
Issue number | 6 |
DOIs | |
State | Published - Aug 1992 |