About the Weighting Factors

Back to Main Page

Since the release of BOG version 1.00, a number of users have asked about the size of the weighting factors used in the Merit Function or MF (see here for a detailed description of the Merit Function). In particular, some people were disturbed by the fact that, for some MF targets, the weighting factors must be larger than for others in order to achieve a "balance" of effects. A typical example, shown in the Advanced Tutorial, involves the balancing of ISO brightness and tilt performance. In that case, an ISO weight of 1 and a Tilt Performance weight of 35 were used. (See here for a clarification of the Tilt Performance target.)

Why not just rescale the various components so that equal weighting factors produce equal contributions?

I experimented with additional correction factors so that the weights would come out more even, but I was unsatisfied, since the correction factors depend very strongly on the gem material and design. For example, here are the ISO brightness and tilt variation values for a few gems, as calculated with BOG's PlayPen:

Gem and Material
Tilt Variation
Ratio ISO / Tilt Variation
BOGTUT.GEM in Quartz
Smith Bar in Quartz

As the table demonstrates, the relative size of ISO brightness and tilt variation can change enormously for different gem materials and designs. It would be impossible to assign appropriate "fudge factors" to make things come out even in all instances. In the end, I found it easier to understand and use the "raw" values. I know what ISO 75% and tilt variation 9% mean, and can then balance them explicitly.

In practice, I use the Evaluate MF button (see here) in the Specify Optimization Targets window to evaluate the current balance of weighting factors. A couple of trials (and maybe a minute or two) is usually enough to get things right.

Back to Main Page

Documentation maintained by Tom Herbst. Last modified 16-Aug-2005