[03:51] MSG: Ping timeout: 245 seconds [05:54] MSG: Read error: Operation timed out [06:58] Join: fiveop joined #corewars [15:12] Join: alexander joined #corewars [15:25] MSG: [15:28] Join: johnkw joined #corewars [16:19] MSG: Read error: Operation timed out [16:20] Join: johnkw joined #corewars [16:21] MSG: Client Quit [16:23] Join: johnkw joined #corewars [16:49] Join: CoreOld joined #corewars [17:02] MSG: Quit: Trillian (http://www.ceruleanstudios.com [18:14] Join: Metcalf joined #corewars [18:14] Hi :-) [18:16] heya met. [18:16] how goes? [18:16] Hi Bvowk [18:16] Okay thanks [18:17] How for you? [18:17] Busy weekend planned? [18:24] MSG: Quit: humhum [18:34] MSG: Ping timeout: 245 seconds [18:34] Join: Metcalf joined #corewars [18:40] Join: OoS joined #corewars [18:42] Join: willvarfa joined #corewars [18:42] evening all [18:42] Hi Will [18:43] h will [18:43] how goes things? [18:43] recmath, for instance? :-) [18:44] Last time I checked I was slightly ahead of Roy :-) [18:44] I'm writing a program to search for entries now [18:45] MSG: Ping timeout: 245 seconds [18:51] last time I checked, Roy was slightly ahead :-) [18:51] damn brilliant results both of you, though [18:51] who else is taking part from the cw world? [18:52] No-one yet, but I'm hoping a few more will [18:59] * willvarfa has written some code, but not run much yet [19:12] Join: fiveop joined #corewars [19:12] Join: yoR joined #corewars [19:13] I think more will join soon ;-) [19:13] As long as we keep bragging about it [19:13] (hi btw) [19:13] Hi yoR [19:14] Last time I checked I was slightly ahead JM ;-) [19:14] Hmmm... we'll see about that [19:14] But you still beat the crap out of me with the higher N's [19:15] so you're doing it by hand john? [19:15] I'm hoping to do better when I've written a program to find solutions for the higher N's [19:15] and roy, you've got a box crunch it? [19:15] By hand for N >= 9 [19:16] I've crunshed the lower N's, up to 8 by computer, the others are handcrafted, but with a lot of help from JM (his tips doubled my GA-scores) [19:17] Btw JM: I've now got a perfect score for 3,4 and 6 (not for 5 yet, which is strange..) [19:20] I only have a perfect score for 3 and 4 [19:21] i have a perfect score for N=OVER 9000 [19:33] you're using genetic algorithms Roy? [19:41] Yup [19:42] Probably not the best way, but it got me perfect scores for the lower ones [19:45] did it take much time and boxes? [19:47] Just one box and maybe 2 days cpu time (for all my calculations) [19:54] genetic algos to choose the constants themselves? [19:58] Yeah, the N's are chosen semi-random (based on probable succes rate) and then put in a huge pool of other N-combo's and the best ones mate and generate other n-combinations, the score is the amount of produced primes [19:59] seems like you could save yourself a lot of work [19:59] i could just be mistaken though [19:59] i have some ideas but i doubt i'll end up coding anything [20:00] Well, tell me (and the others) we might use it :) [20:02] haha, but do i get a cut of the prize!? [20:02] ;) [20:03] You want half the statue? [20:03] yeah, lets just cut it up [20:03] ;P [20:03] nah, i'm just joking [20:03] knowing myself i won't get anywhere serious so sure [20:03] Will still look intresting probably haha [20:03] yeah, i want to buy some now that i saw em [20:04] 1) metcalf already determined that the greatest number of odd combinations will come from one odd number and the rest even numbers [20:04] 2) this means that for best results, we want all the even numbers to come out so that every combination of the even numbers ends in a value such that [20:04] value+odd and value-odd are prime [20:05] first step: find prime pairs for odd numbers [20:05] randomly or sequentially or whatever, pick an odd number, double it, and test primes to see if prime+N is also prime [20:05] the largest pool of values for a given step is perhaps your best bet to start with [20:05] * willvarfa went that way, ish [20:06] then find gaps between the 'target' values [20:06] (the ones in between the two primes) [20:06] and start looking for, say, [20:06] situations where two gap sizes also have values such that their combinations lead to primes [20:06] i suppose you could abstract that to any quantity of levels, but it might be simpler and quicker to first find constants one or two levels deep [20:06] * evitable shrugs [20:07] i haven't done anything so i don't know how likely that is to work out, but it seems to me that there are probably a lot of primes separated by say, 2 [20:07] er, 3 [20:07] but there may be a lot separated by large numbers as well [20:08] where i was losing my way was abstracting the combinations of constants.. i guess you could just keep measuring gaps and looking for the values with the most gaps of a given size [20:08] since prime numbers are mathematically special though, i wouldn't be surprised if it turns out this is a majorly poor strategy because of some inherent property i'm unaware of :) [20:11] anyway, that's what i mean by saving yourself some work.. you can probably narrow down the values you are selecting from and improve the effectiveness of your genetic thingy [20:11] I tried something a little similair, your first step anyway, but then I gave up because the results where a bit surprising [20:11] Later (thanks to JM) I found that the surprising bit was a good thing, I just didn't know what to do with it [20:13] :-) [20:15] * willvarfa is biting lip [20:16] joonas has explained a few things to me after I went to him scratching my head about the results of that approach :-) [20:18] heh [20:19] define "perfect score", Roy [20:19] means max possible values [20:19] i would think primes = the number of combinations in (N-1) [20:19] yeah, but by my logic, I'm not finding anything close [20:19] would be a perfect score [20:19] and if you have N=6, how do you not have an N=5 and N=4 nested within it? [20:27] Maybe I have, not idea :) [20:27] Odds and even N's are different [20:27] Remember that some combinations of odd+odd etc are different [20:27] But I'm off, bye! [20:28] * OoS waves [20:28] The maximum for N=3 for example of 9, and you can explain it by writing down all the combinations (there are more then 9 combinations possible) [20:28] of>is [20:28] MSG: [20:32] * willvarfa waves [20:32] hmm [20:32] maximum is 9? [20:32] Yes [20:34] but many will be negative [20:34] and one must be odd, meaning that two are even and cannot be prime [20:34] etc [20:34] The negative values all have a positive equivalent [20:34] a - b - c [20:35] e.g. if your values are a, b and c [20:35] where b and c are larger than a [20:35] is not a prime [20:35] hmm, I'm learning a lot more about maths than about code from this :-) [20:36] if a is your odd number, you only need to check a, a+b, a+c, a+b+c, a+b-c, a-b+c, a-b-c, a-b, a-c [20:36] Taking the absolute values of course [20:38] makes sense [20:39] It's pretty easy to knock up a program to find a solution for N=3 [20:55] hrm. [21:00] heh now you guys got me thinking again [21:00] here's what i pondered at lunch [21:01] you can have one step selecting random(?) odd numbers and making a count/list of how many prime pairs there are for that step [21:01] that's the goal, then; you want even constants that in all combinations produce numbers from that list [21:02] i suppose you could rank them by quantity, i.e. the most pairs is the one to pay the most attention to, but i have no proof that that's the best way, so i think devoting some time to other sets might be good, or even promoting them in priority if they produce better results? [21:02] anyway, then i was thinking, for each constant, the values to test are every combination of the previous constants +- the new constant [21:03] Why is everyone else recommending a top down approach [21:03] ? [21:03] Am I the only one with a bottom up approach. Have I missed something? :-( [21:03] just made the most sense at the moment, i'm talking it out [21:03] I feel like the odd one out :-) [21:03] so from the target list, you pick two constants and test their combinations [21:04] if you only want to accept a perfect score, you can then keep picking constants and checking [21:04] building up your list fo constants and each time picking a new one and trying it vs the most likely suspects [21:04] you don't have to iterate any further than that i think [21:05] each time you find a perfect set, you can store all the combos that were produced by those constants and then when you test a new one to add to it, you just test +-N on each item [21:05] does that sound smart or stupid? [21:05] it probably leaves couple numbers unfound [21:05] how so? [21:06] for a perfect result, every step must be perfect [21:06] maybe there isn't a perfect solution though for certain values of N [21:06] i have a hunch, but OTOH i am not much of a mathematician [21:06] anyway, it's possible to get any valid numbers the way i described.. the thing in question is how valuable each set is [21:06] naiively it would seem that the larger sets are better bets [21:07] since they have more possible combinations [21:07] There aren't perfect solutions for N=4+ [21:07] how do you know? :) [21:07] also yoR has a perfect score for 6 [21:07] he said earlier [21:07] anyway, you don't HAVE to aim for perfect the way i described, i guess [21:08] Perfect score <> all combinations are prime [21:08] Perfect score = maximum number of combinations are prime [21:08] right [21:09] don't know what that's supposed to mean [21:09] :P [21:10] heh, i don't know any language well except mirc scripting D: [21:10] maybe i ought to play with something over the weekend [21:10] i feel like it now but i bet i won't when i get home and can do it [21:11] You can try making solutions by hand and submitting them :-) [21:13] i can indeed [21:13] a human mind behind it may very well be the best thing [21:13] but i want to see some stuff before i even try [21:14] there's an example on the task description on the website [21:16] oh i remember the other thing i was going to say [21:16] miz was talking about 'breeding' his numbers [21:16] but i wonder what that entails? [21:16] i can't immediately think of anything to do with a set of effective numbers to generate other likely effective numbers [21:16] :\ [21:16] then again i only have like up to pre-calculus in public high school [21:17] and i don't rember anythin past algebra anymore, if i'm lucky [21:24] you're lucky [21:26] pff, i liked math [21:26] not going to college wasn't lucky, it was stupid [21:30] * OoS was stupid too! [21:33] serious? o_O [21:48] I'm planning a little surprise for when yoR looks at the standings tomorrow ;-) [22:05] hrm. [22:18] * willvarfa waves [22:18] MSG: Quit: Ex-Chat [22:34] Join: CoreOld joined #corewars [22:34] Hi CoreOld [22:35] hey [22:35] :-) [22:35] back from cinema [22:35] dark knight... [22:35] What did you think of it? [22:36] Haven't seen batman begins ... so I don't know if I got all hints [22:36] but it's cool [22:36] heath ledger plays the joker really well [22:36] I liked it [22:36] so...evil ;-) [23:33] MSG: Quit: Trillian (http://www.ceruleanstudios.com [23:45] wow there's lots of pairs