<BODY bgcolor="#aaeebb"> <CENTER> <H3> Principle of Ordinal Optimization Applications<BR> <BR> Original Problem ............ Surrogate Problem<BR> (hard) ............................... (easier)<BR> </H3> <H4> Steady State Simulation ....................... Shorter Simulation<BR> Many Samples Replications ............. Fewer Samples Replications<BR> Rare Events ................................... Not-so-rare Events<BR> High Threshold ................................... Tight Threshold<BR> Complex Model ....................................... Simple Model<BR> </H4> </CENTER> <BR> <tt> <A HREF="Slide14.html" target="_parent">To Next Slide</A> / <A HREF="Slide12.html" target="_parent">To Previous Slide</A> / <A HREF="../OOTOC.html" target="_parent">Table of Content</A> </tt> <BR> <hr> <hr> <br> <P> In fact, if we summarize the application of OO so far in the literature (see the annotated reference list), then this slide summarizes the principle. Basically, we use a very crude surrogate model to predict the relative rank order of various design choices. While these surrogate models estimate the performance value very poorly, they can be very good in estimating order, particularly if we soften the goal of order alignment. Note we use the term &quot;surrogate model&quot; in a very broad sense including operating the system in question under different environments, e.g., to estimate the rank of highly reliable systems using accelerated aging or adverse operating conditions in order to save time. <BR> <BR> In this sense, all engineering analysis is based on the same principle. The only point we are emphasizing here is that for some purposes the models need not be as precise as we are taught to strive for.<BR> <BR> </BODY> </HTML>