Abstract | ||
---|---|---|
We introduce a real-world job shop scheduling problem where the objective is to minimize configuration costs that depend on the sliding pairwise similarity between two assets ordered one after the other in a processing batch. This implies that our fundamental challenge is to learn from the costs what constitutes asset similarity in the context of batching locally and optimizing a multi-line end to end. We present a 3 component scheduling system: simulator, scheduler and hyper-optimizer. The scheduler relies upon a machine learning algorithm -- hierarchical clustering, to select, from an entry yard, assets for a batch based on weighted similarity. It then utilizes a weighted distance matrix to sequence the assets. The weights used by the scheduler are optimized online with an evolutionary algorithm. |
Year | DOI | Venue |
---|---|---|
2016 | 10.1145/2908961.2931647 | GECCO (Companion) |
Keywords | Field | DocType |
hyper-optimization, machine learning | Hierarchical clustering,Pairwise comparison,Fixed-priority pre-emptive scheduling,Mathematical optimization,Evolutionary algorithm,End-to-end principle,Scheduling (computing),Matrix (mathematics),Computer science,Job scheduler,Artificial intelligence,Machine learning | Conference |
Citations | PageRank | References |
0 | 0.34 | 6 |
Authors | ||
3 |
Name | Order | Citations | PageRank |
---|---|---|---|
Ignacio Arnaldo | 1 | 81 | 7.69 |
Erik Hemberg | 2 | 143 | 35.68 |
Una-May O'Reilly | 3 | 1477 | 181.38 |