305. An algorithm selection approach for the flexible job shop scheduling problem: Choosing constraint programming solvers through machine learning
Invited abstract in session MB-7: Advances in scheduling: From theory to practice, stream Scheduling and Project Management.
Monday, 10:30-12:00Room: Clarendon GR.01
Authors (first author is the speaker)
| 1. | David Müller
|
| Faculty of Computer Science, Nuremberg Institute of Technology Georg Simon Ohm | |
| 2. | Marcus G. Müller
|
| Institute of Robotics and Mechatronics, German Aerospace Center (DLR) | |
| 3. | Dominik Kress
|
| Helmut Schmidt University | |
| 4. | Erwin Pesch
|
| Faculty III, University of Siegen |
Abstract
Constraint programming solvers are known to perform remarkably well for most scheduling problems. However, when comparing the performance of different available solvers, there is usually no clear winner over all relevant problem instances. This gives rise to the question of how to select a promising solver when knowing the concrete instance to be solved. We aim to provide insights into this question for the flexible job shop scheduling problem. First, we investigate relative performance differences among five constraint programming solvers on various problem instances. We find that two solvers, the IBM ILOG CPLEX CP Optimizer and Google’s OR-Tools, outperform alternative solvers. Both solvers show complementary strengths regarding their ability to determine provably optimal solutions within practically reasonable time limits and their ability to quickly determine high quality feasible solutions across different test instances. Hence, we leverage the resulting performance complementarity by proposing algorithm selection approaches that predict the best solver for a given problem instance based on instance features. The approaches are based on two machine learning techniques, decision trees and deep neural networks, in various variants. In a computational study, we analyze the performance of the resulting algorithm selection models and show that our approaches outperform the use of a single solver and should be considered as a relevant tool by decision makers in practice.
Keywords
- Scheduling
- Machine Learning
Status: accepted
Back to the list of papers