Deutsch
 
Hilfe Datenschutzhinweis Impressum
  DetailsucheBrowse

Datensatz

DATENSATZ AKTIONENEXPORT

Freigegeben

Forschungspapier

An Integer Interior Point Method for Min-Cost Flow Using Arc Contractions and Deletions

MPG-Autoren
/persons/resource/persons118120

Becker,  Ruben
Algorithms and Complexity, MPI for Informatics, Max Planck Society;

/persons/resource/persons44737

Karrenbauer,  Andreas
Algorithms and Complexity, MPI for Informatics, Max Planck Society;

/persons/resource/persons45021

Mehlhorn,  Kurt
Algorithms and Complexity, MPI for Informatics, Max Planck Society;

Externe Ressourcen
Es sind keine externen Ressourcen hinterlegt
Volltexte (beschränkter Zugriff)
Für Ihren IP-Bereich sind aktuell keine Volltexte freigegeben.
Volltexte (frei zugänglich)

arXiv:1612.04689.pdf
(Preprint), 278KB

Ergänzendes Material (frei zugänglich)
Es sind keine frei zugänglichen Ergänzenden Materialien verfügbar
Zitation

Becker, R., Karrenbauer, A., & Mehlhorn, K. (2016). An Integer Interior Point Method for Min-Cost Flow Using Arc Contractions and Deletions. Retrieved from http://arxiv.org/abs/1612.04689.


Zitierlink: https://hdl.handle.net/11858/00-001M-0000-002C-5714-E
Zusammenfassung
We present an interior point method for the min-cost flow problem that uses arc contractions and deletions to steer clear from the boundary of the polytope when path-following methods come too close. We obtain a randomized algorithm running in expected $\tilde O( m^{3/2} )$ time that only visits integer lattice points in the vicinity of the central path of the polytope. This enables us to use integer arithmetic like classical combinatorial algorithms typically do. We provide explicit bounds on the size of the numbers that appear during all computations. By presenting an integer arithmetic interior point algorithm we avoid the tediousness of floating point error analysis and achieve a method that is guaranteed to be free of any numerical issues. We thereby eliminate one of the drawbacks of numerical methods in contrast to combinatorial min-cost flow algorithms that still yield the most efficient implementations in practice, despite their inferior worst-case time complexity.