Scheduling a Scheduling Competition Providence Sept 2007 Commentary

  • Slides: 8
Download presentation
Scheduling a Scheduling Competition, Providence, Sept 2007 Commentary on Session A J. Christopher Beck

Scheduling a Scheduling Competition, Providence, Sept 2007 Commentary on Session A J. Christopher Beck Department of Mechanical & Industrial Engineering University of Toronto, Canada jcb@mie. utoronto. ca © J. Christopher Beck 2007

1+3 Papers l Ghersi et al. focuses on an operational question – how do

1+3 Papers l Ghersi et al. focuses on an operational question – how do we judge the results? – important, technical issue – valuable for whoever runs the competition l Other 3 papers focus on a more strategic question – what problem types should the competition address? 2 © J. Christopher Beck 2007

Some Operational Issues l Automated verification of results l Entries need to run on

Some Operational Issues l Automated verification of results l Entries need to run on the same platform l Not just source code vs. binary – License issues? 3 © J. Christopher Beck 2007

Integration vs. Focus l Rich problems (Le Pape; Guerri et al. ) vs. a

Integration vs. Focus l Rich problems (Le Pape; Guerri et al. ) vs. a single fundamental issue (Cicirello) l Things to consider: – industry impact – potential for research progress/breakthroughs • will we understand the results? – barriers to entry • easier to participate in Cicirello-style track – is this an OR vs. AI issue? 4 © J. Christopher Beck 2007

Multiple Tracks? l Perhaps of increasing difficulty l Things to consider: – spreading the

Multiple Tracks? l Perhaps of increasing difficulty l Things to consider: – spreading the competition too much • one entry per track is not very interesting – “granularity” of tracks – creating a challenge – barriers to entry 5 © J. Christopher Beck 2007

Robustness as a Criteria l Bias evaluation toward good performance on all instances (Le

Robustness as a Criteria l Bias evaluation toward good performance on all instances (Le Pape) l Best all-round single machine scheduler across different opt. funcs (Cicirello) l Things to consider: – “jack of all trades, master of none” • give up on being “the” best on a given problem • marketing – industry vs. research – another OR vs. AI issue? 6 © J. Christopher Beck 2007

Competition or “Challenge” l Competition – like SAT or Planning competition – multiple tracks

Competition or “Challenge” l Competition – like SAT or Planning competition – multiple tracks l Challenge – one problem type (e. g. , one of Guerri et al. ’s, one of Le Pape’s) – long work horizon (e. g. , 6 months – 1 year) – like the CP Modeling Challenge (2005) 7 © J. Christopher Beck 2007

Competition or “Challenge” l Things to consider: – marketing – (end user) industry interest

Competition or “Challenge” l Things to consider: – marketing – (end user) industry interest and commitment – barriers to entry – potential for lack of community interest – organizational overhead 8 © J. Christopher Beck 2007

Providence Esther Providence Introduction Providence Introduction We dontProvidence Esther Providence Introduction Providence Introduction We dont
Adding Commentary Commentary Commentary is relating your detailAdding Commentary Commentary Commentary is relating your detail
Social Commentary Subtitle Social Commentary Social Commentary isSocial Commentary Subtitle Social Commentary Social Commentary is
Ineffective commentary Commentary that repeats Commentary that isIneffective commentary Commentary that repeats Commentary that is
Imperfect Competition Imperfect Competition Monopolistic Competition Monopolistic competitionImperfect Competition Imperfect Competition Monopolistic Competition Monopolistic competition
Providence of God Providence of God Foresight andProvidence of God Providence of God Foresight and
PROVIDENCE UNIVERSITY Providence University Your Gateway to thePROVIDENCE UNIVERSITY Providence University Your Gateway to the