For Discrete Event Simulation Under Ordinal Optimization
- The probability that nothing good (i.e., element of G) is in the
selected set S decreases "exponentially" fast with increasing
G and S.
- The perturbing noise (i.e., errors that will push G outside of
S) decreases with increasing length of simulation
- These two effects MULTIPLY ==> many orders of magnitude improvement
in convergence.
Experimentally verified on many examples so far and now with
theoretical justification.
To Next Slide
/
To Previous Slide
/
Table of Content