Abstract
This paper presents a mean squared error analysis of the harmonic gradient estimators for steady-state discrete-event simulation outputs. Optimal mean squared errors for the harmonic gradient estimators are shown to converge to zero as the simulation run length approaches infinity at the same rate as the optimal mean squared errors for the symmetric (two-sided) finite-difference gradient estimator. Implications of this result are discussed.
| Original language | English (US) |
|---|---|
| Pages (from-to) | 573-590 |
| Number of pages | 18 |
| Journal | Journal of Optimization Theory and Applications |
| Volume | 80 |
| Issue number | 3 |
| DOIs | |
| State | Published - Mar 1994 |
| Externally published | Yes |
Keywords
- Simulation
- frequency domain methodology
- gradient estimation
- harmonic analysis
ASJC Scopus subject areas
- Control and Optimization
- Management Science and Operations Research
- Applied Mathematics