I have a very long running process that is testing randomly generated scenarios. At any given point in time I have two knowns: $T is the total number run so far; and $S is the number of those that have been successful. From those I can easily indicate the current success rate as a percentage.
I'm looking for some mechanism to indicate whether the processing is nearing -- or tending towards -- some level of completion.
If I introduce a third variable, target success rate, $TSR (say: 90%), then what further statistics can I derive that would give me a feel for how the process is progressing?
For example, it would be possible to calculate a number of further tests that would need to be run, that -- assuming they were all successful -- would raise the success rate to the target value. But assuming the current success rate is less than the $TSR -- otherwise I wouldn't be posting -- that would be a naive statistic.
So then I thought about trying to calculate the number of addition tests required to achieve the $TSR, assuming the current success rate continued; but of course, that will never happen.
So then I thought about trying to calculate the trend in the current success rate; and if that trend shows a consistent decline, or reaches a point where the $TSR can never be reached, then I abandon processing with the current parameters and move on to the next set.
I don't have any code because I'm just starting to think about the problem; and so far, I have got a clue what to code.
If any of this sparks any thoughts, possibilities or suggestions; please post.
In reply to Instrumenting a genetic algorithm. by BrowserUk
| For: | Use: | ||
| & | & | ||
| < | < | ||
| > | > | ||
| [ | [ | ||
| ] | ] |