Heuristic solutions can be improved.
You're still missing the point. You can only know when to stop looking for an improvement (either through changes to the algorithm or further iterations), if you have some target to aim for. Maybe 446 is the best for this particular dataset. Or maybe, there is another 2 point gain to be had? Or a 20 point gain? Or 200? Maybe there are hundreds of better solutions to this particular dataset. Without running an exhaustive search you have no way of judging how good any particular heuristic is for any given dataset.
And whilst your modifications find better solutions for this dataset, maybe my original will work better for a different dataset? Or many different datasets. And without an exhaustive search of a representative sample of datasets, you've no way to compare the efficacy of different heuristics.
In reply to Re^5: Challenge: Letter Power
by BrowserUk
in thread Challenge: Letter Power
by Tanktalus
| For: | Use: | ||
| & | & | ||
| < | < | ||
| > | > | ||
| [ | [ | ||
| ] | ] |