Efficient Suboptimal Rare Event Simulation (with X. Zhang and P. Glynn).

Summary:

This paper considers algorithms that are not optimal under traditional definitions of optimality in rare event simulation (such as asymptotic optimality or strong efficiency), yet the algorithms are shown to be a substantial improvement upon crude Monte Carlo. The motivation is that often it is very difficult to construct optimal algorithms. We illustrate, for instance in a heavy-tailed setting, that an easy algorithm obtained out of conditioning is already a big improvement. The paper also includes a Lyapunov criterion for lower bounds.

Bibtex:

@INPROCEEDINGS {ZBG07,

AUTHOR={ X. Zhang and J. Blanchet and P. Glynn },

YEAR={2007},

TITLE={ Efficient suboptimal rare event simulation },

BOOKTITLE={Proceedings of the 2007 Winter Simulation Conference},

EDITOR={ S. G. Henderson, B. Biller, M.-H. Hsieh, J. Shortle, J. D. Tew and R. R. Barton },

PUBLISHER={IEEE Press},

PAGES={607-614}

}