Large Deviations Perspective on Ordinal Optimization of Heavy-tailed Systems (with J. C. Liu and B. Zwart)

Summary:

We assume that there are d systems, each of which can be independently simulated and for each of the simulations we can evaluate the performance of the corresponding system and obtain a score. Given a total of n replications we are interested in budget allocation to select the system with the smallest mean score to minimize the probability of false detection (PFS). We use large deviations theory and improve upon results in Broadie, Han and Zeevi (WSC 2007).

Bibtex:

@INPROCEEDINGS {BlanLiuZwart10,

AUTHOR={ J. Blanchet and J. C. Liu and B. Zwart},

YEAR={2008},

TITLE={Large deviations perspective on ordinal optimization of heavy-tailed systems},

BOOKTITLE={Proceedings of the 2008 Winter Simulation Conference},

EDITOR={S. J. Mason, R. R. Hill, L. Moench and O. Rose},

PUBLISHER={IEEE Press},

PAGES={489-494}

}