1 / 8

Hardness analysis

Hardness analysis. Tommy Messelis, Patrick De Causmaecker. Intro. Empirical Harndess Models model the empirical hardness performance of some algorithm(s) on a specific instance as a function of features computationally inexpensive ‘properties’ of the instance

neila
Download Presentation

Hardness analysis

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Hardness analysis Tommy Messelis, Patrick De Causmaecker

  2. Intro • Empirical Harndess Models • model the empirical hardness • performance of some algorithm(s) on a specific instance • as a function of features • computationally inexpensive ‘properties’ of the instance • allow for performance prediction • algorithm portfolio approach • parameter tuning

  3. Procedure • introduced by K. Leyton-Brown et al. • Identify a problem instance distribution • Selectone or morealgorithms • Choose a feature set • Generate an instance set from the distribution, calculate all features and determine all performance criteria • Eliminateredundant/uninformative features • Use machine learning techniques to select functions of the features that approximate the algorithm’s performance(s) K. Leyton-Brown, E. Nudelman, Y. Shoham. Learning the empirical hardness of optimisation problems: The case of combinatorial auctions. In LNCS, 2002

  4. So far • successful in: • Winner Determination Problem (combinatorial auctions) • Propositional Satisfiability Problem (SAT) • Scheduling & Timetabling • Nurse Rostering Problem (at least on a small scale) • directly on NRP representation • on a translation to SAT • slightly more accurate models !

  5. Conclusions • Translating to SAT is very usefull • abstract way of thinking • no expert knowledge needed • there is already an extensive set of features • what can we do for other problems?

  6. What is to come • MaxSAT • optimisation variant of SAT problems • literature only contains runtime predictions for SAT • predict other performance criteria • for other algorithms aiming to find an optimal solution

  7. What is to come • about translation of NRP instances into SAT • optimal solution for the resulting SAT problem is not the optimal solution for the NRP problem • one constraint generates an arbitrary number of clauses • there is no concept that keeps clauses together • no conservation of the objective function of the NRP instance • groupSAT as a new model • keep clauses together • adapt maxSAT algorithms • in the end: solve NRP by (adapted) state-of-the-art maxSAT algorithms

  8. Questions?

More Related