40 likes | 152 Vues
This wrap-up paper presents critical insights from three key papers on scheduling benchmarks, including reactive, over-subscribed, and real-life scheduling. It also reflects on experiences from international competitions like the International Timetabling Competition and the International Planning Competition. Key discussions highlight the need to learn from diverse approaches, the significance of bridging theory and practice, and the value of adapting benchmarks to real-life scenarios. Overall, the findings encourage the advancement and promotion of scheduling research while evaluating the effectiveness of current systems and algorithms.
E N D
Session B Wrap-upSummary and Commentary Roman Barták Charles University (Czech Republic) roman.bartak@mff.cuni.czhttp://ktiml.mff.cuni.cz/~bartak
Summary • Three papers proposing benchmark areas • Reactive scheduling • Over-subscribed scheduling • Real-life scheduling • Wrap-up paper on benchmarks • Two papers describing experience with other competitions • International Timetabling Competition • International Planning Competition SSC 2007, Roman Barták
Other competitions • We should learn from others. • We are completely different. • Why should one run a (scheduling) competition? • compare the systems or algorithms? • bridging the gap between theory and practice? • advance the research? • drive the research? • promote the research? • just to play? • … SSC 2007, Roman Barták
Benchmarks • real-life motivated vs. academic problems • going beyond classical scheduling • dynamic aspects • over-subscribed • different objectives (robustness,…) • common representation • language for covering „all“ scheduling problems • generating problem instances vs. real-life data • parameters to vary problem hardness or to vary problem type • evaluation criteria • is runtime so bad? • conclusions from the competition • any winner? SSC 2007, Roman Barták