II. SIMPLE EXPLANATIONS OF KEY IDEAS IN ORDINAL OPTIMIZATION
II.1. THE WHAT AND WHY OF ORDINAL OPTIMIZATION IN LESS THAN 200 WORDS
BY Y.C. HO
- The rise of human-made problems (manufacturing, communication, C3I,
etc.) where traditional real-variable based analysis are less
applicable, and the new developments in hardware (parallel computing
and cheap memory) and software (OOP and component-ware) dictate new
paradigm for design and performance optimization.
- While Simulation is the only general performance tool of choice, it is
not satisfactory for many of these problems. The main reason is the
time consuming runs, especially when parametric search for good
designs are involved with a combinatorially large search space.
There are fundamental limitations to improve the simulation speed due
to fact that confidence interval of performance estimate decreases at
best at the rate of 1/n1/2 where n denotes
the length of simulation.
- OO is an attempt to circumvent these difficulties at least for the
initial stages of optimization search. It rests on two ideas: (i)
Order is much easier to determine than value --- It is easier to
determine if A > B than to determine A - B = ? (ii)
Softening the goal makes hard problem easier --- Instead of asking the
best for sure let us settle for the good enough with high probability.
Often times several orders of magnitude speed up can be achieved in
simulation experiemnts when we adopt these two viewpoints.
To Next Slide
/
To Table of Content