Loading...
  OR  Zero-K Name:    Password:   

Post edit history

What's the difference between a +10 mex and 5 +2 mexes?

To display differences between versions, select one or more edits in the list using checkboxes and click "diff selected"
Post edit history
Date Editor Before After
4/21/2014 3:29:34 PMGBrankTheEloIsALie before revert after revert
4/21/2014 3:27:55 PMGBrankTheEloIsALie before revert after revert
4/21/2014 3:26:44 PMGBrankTheEloIsALie before revert after revert
4/21/2014 3:26:15 PMGBrankTheEloIsALie before revert after revert
4/21/2014 3:24:21 PMGBrankTheEloIsALie before revert after revert
4/21/2014 3:17:49 PMGBrankTheEloIsALie before revert after revert
4/21/2014 3:16:42 PMGBrankTheEloIsALie before revert after revert
4/21/2014 3:15:53 PMGBrankTheEloIsALie before revert after revert
4/21/2014 3:14:17 PMGBrankTheEloIsALie before revert after revert
Before After
1 Well, here it is! 1 Well, here it is!
2 \n 2 \n
3 Computationally inexpensive, straightforward and so far passing all examples :D 3 Computationally inexpensive, straightforward and so far passing all examples :D
4 \n 4 \n
5 (I'll spare the derivation of these formulas, I already outlined it above). 5 (I'll spare the derivation of these formulas, I already outlined it above).
6 \n 6 \n
7 Input: [i]n[/i] mexes, mex incomes [i]m_i[/i], energy available for OD [i]E[/i] 7 Input: [i]n[/i] mexes, mex incomes [i]m_i[/i], energy available for OD [i]E[/i]
8 Output: Allocated energies per mex [i]e_i[/i] 8 Output: Allocated energies per mex [i]e_i[/i]
9 \n 9 \n
10 1. Calculate k = 1/(4 sqrt(E + 4 n)) * sqrt(sum(m_i²)) 10 1. Calculate k = 1/(4 sqrt(E + 4 n)) * sqrt(sum(m_i²))
11 2. For each mex, set e_i = m_i²/(16 k²) - 4 11 2. For each mex, set e_i = m_i²/(16 k²) - 4
12 \n 12 \n
13 That's basically it. 13 That's basically it.
14 \n 14 \n
15 Exceptions: 15 Exceptions:
16 -This formula may result in negative e values, in which case the respective e_i should be set to 0 and the calculation restarted (this time omitting that particular mex that is not worth overdriving). 16 -This formula may result in negative e values, in which case the respective e_i should be set to 0 and the calculation restarted (this time omitting that particular mex that is not worth overdriving).
17 -If any grid doesn't hold enough e for the requested amounts, solve the problem for the grid locally (with E = the max e available in the grid) and redo the calculation for the remaining stuff (with the remaining energy). (This is basically what the current algorithm already does). 17 -If any grid doesn't hold enough e for the requested amounts, solve the problem for the grid locally (with E = the max e available in the grid) and redo the calculation for the remaining stuff (with the remaining energy). (This is basically what the current algorithm already does).
18 \n 18 \n
19 It will make sense to calculate the e_i beginning with the [i]smallest[/i] mexes, so that negative values can immediately be ruled out. k will have to be recalculated each time, but that's acceptable (applying that formula even for every mex should really not be a problem if you have all the m_i at hand). 19 It will make sense to calculate the e_i beginning with the [i]smallest[/i] mexes, so that negative values can immediately be ruled out. k will have to be recalculated each time, but that's acceptable (applying that formula even for every mex should really not be a problem if you have all the m_i at hand).
20 \n 20 \n
21 2 examples (2 mexes, 3 mexes) of cases without the exceptions: 21 2 examples (2 mexes, 3 mexes) of cases without the exceptions:
22 http://pastebin.com/b3AjS5HD 22 http://pastebin.com/b3AjS5HD
23 \n 23 \n
24 No more manual grid color balancing :) 24 No more manual grid color balancing :)
25 \n 25 \n
26 PS: It's sort of odd to notice that the current gadged is actually quite close, seeing as it uses energy proportional to m_i², too. It's just ( sort of) missing the constant term there. Well, I really can't get a decent interpretation out of that formula atm ( although i notice now that there's some cancellation I missed, the 16 in the second formula is pretty unnecessary if you just cancel it with the first 4 in the first formula) . 26 PS: It's sort of odd to notice that the current gadged is actually quite close, seeing as it uses energy proportional to m_i², too. It's just ( sort of) missing the constant term there. Well, I really can't get a decent interpretation out of that formula atm ( although i notice now that there's some cancellation I missed, the 16 in the second formula is pretty unnecessary if you just cancel it with the first 4 in the first formula. I'd just introduce mistakes if I'd attempt to change it now) .
27 Also note that I have no proof for this formula thus far. The "equal derivatives" was just sort of empirical from 3 examples, please tell me if you manage to find a counter-example! (SCIENCE!) 27 Also note that I have no proof for this formula thus far. The "equal derivatives" was just sort of empirical from 3 examples, please tell me if you manage to find a counter-example! (SCIENCE!)
28 \n 28 \n
29 Well, that felt good. I need to do this stuff more often :P 29 Well, that felt good. I need to do this stuff more often :P