[Home]History of SplitTargeting

Robo Home | Changes | Preferences | AllPages

Revision 8 . . September 23, 2007 22:34 EST by AaronR [I think I tried that]
Revision 7 . . September 23, 2007 22:02 EST by Skilgannon [my take]
  
Difference (from prior major revision) (no other diffs)

Changed: 13c13,14
When I was reading the DynamicSegmentation? page, I thought of something that seems like the name SplitTargeting would fit best. Basically, you look at all GF's greater than the 'middle', and all less, and you decide which has a higher value. Then you look at all GF's greater than the 'middle' of the option you chose, and all less, etc, until you are down to one bin. That's the bin you fire at. You choose the middle as the position they are at, and max and min would have to be precise-predicted, otherwise you would almost never shoot at negative GFs. This method should not shoot at the highest point in the graph, but instead do a sort of binary search for the densest part of the curve. -- Skilgannon
When I was reading the DynamicSegmentation? page, I thought of something that seems like the name SplitTargeting would fit best. Basically, you look at all GF's greater than the 'middle', and all less, and you decide which has a higher value. Then you look at all GF's greater than the 'middle' of the option you chose, and all less, etc, until you are down to one bin. That's the bin you fire at. You choose the middle as the position they are at, and max and min would have to be precise-predicted, otherwise you would almost never shoot at negative GFs. This method should not shoot at the highest point in the graph, but instead do a sort of binary search for the densest part of the curve. -- Skilgannon
* I tried something like that at one point, it isn't really as effective as just a straight choose-the-highest-bin algorithm. -- AaronR
Robo Home | Changes | Preferences | AllPages
Search: