[an error occurred while processing this directive]
[an error occurred while processing this directive]
|
|
|
Jingxiang Luo Department of Computer Science University of Saskatchewa Saskatoon, Saskatchewan |
|
In performance evaluation, which measures the quality of a stochastic system using some metrics, it is very popular to employ Monte-Carlo simulation. When the metrics depend heavily on some rare events of interest, it is a great challenge to the efficiency of simulation. In order to enhance simulation efficiency, much effort has been put into investigation of this issue recently.
In this talk, first I will give a survey on two mainstream methods of efficient rare event simulation, importance sampling and importance splitting. Both methods have advantages and suffer from curses (failure cases). To be focused, the discussion is based on a class of Markov event systems and some examples arising from queuing scenario.
Then I will present an overview on my recent work, which suggests some novel strategies where the existed methods fail to be efficient. I also highlight some issues that have not received due attention.
Jingxiang Luo is a PhD student at the Dept. of Computer Science, University of Saskatchewan. He is supervised by Prof. Winfried Grassmann. He received a M.Sc. in 2000 from the same department. Previous to that, he held a B.Sc. degree in statistics from CUST (Chinese University of Science and Technology).
His research interest is in studying performance of a computer system, and in stochastic simulation. Current research is in efficient Monto-Carlo simulation when the performance metrics depend heavily on some rare occured events. He is also interested in optimization and operation research, particularly the design of fast algorithms in these areas. [an error occurred while processing this directive]