[00:04] MSG: Quit: Lost terminal [04:09] MSG: Ping timeout: 240 seconds [04:10] Join: Caelian|w joined #corewars [07:32] Join: Nenad joined #corewars [07:47] Join: fiveop joined #corewars [07:49] hi fiveop [07:52] hi [07:53] so, what have you been up to lately? [08:03] MSG: [08:09] MSG: Ping timeout: 240 seconds [09:46] Join: Core_new joined #corewars [10:09] Join: Core_new_ joined #corewars [10:11] MSG: Ping timeout: 240 seconds [12:52] MSG: Read error: Operation timed out [12:56] Join: triple joined #corewars [15:19] MSG: Ping timeout: 240 seconds [15:27] Join: fiveop joined #corewars [16:31] Join: bvowk joined #corewars [16:31] greetings humans. [16:33] <--- tired. [17:01] MSG: Quit: Trillian (http://www.ceruleanstudios.com [18:36] Join: Nenad joined #corewars [18:37] hello ppl [18:39] hN [18:40] heya [18:40] * Nenad streches [18:40] how's it going nenad? [18:40] miz? [18:40] hey! :P [18:40] haven't seen you for months, I think [18:40] hey? [18:40] indeed. [18:40] well, I've never really *seen* you, but... [18:41] well, just recovering from some java coding by reading Fourrier transforms for one of my soon-to-come exams [18:42] I just hope that someone is going to use that Corewar optimizer when it's completely debugged... [18:42] was having a nice evening-nap, and suddenly they bring a pump-truck right in front of our doorway (its 2130), and its seems they just pumped some sewage tank, since the smell forced me to rise up and close the window [18:42] I guess an oficial release will be in max 2 weeks from now [18:43] either that or the dog gave a really smelly fart [18:43] @Miz an average evening, I'd say... stuff like that is known to happen here often [18:44] bvowk: how are your projects? [18:49] not so bad.. [18:50] I should have some more time for them in the next little while :) [18:52] what will be the target this time? [18:52] nano? [18:52] tiny? [18:53] something bigger? [18:53] or not-cw-related? [18:56] heh [18:56] still working on tiny.. [18:57] and I've got a finance project I'm playing with [18:57] just got back from Dublin this morning at like 3AM, so I'm kinda tired. [18:57] oh. how was your trip? [18:58] been testing te quality of Irish beer? [18:58] :) [18:59] anyway, I've been thinking about something... [19:00] oh great, "CorewarOptimizer" is case sensitive [19:00] in you evolvers, did you ever try to optimize values in your warriors in another way [19:00] the trip was alright, getting there and back really sucked. [19:00] Miz: yeah, why? [19:01] @bw - like, generate the instruction genotype and then try to get as good values for it as you can in a short period of time [19:01] Nenad: annoying [19:01] @bw - and then proceed further with the process [19:01] Nenad: yeah, I've tried a whole pile different optimization techniques.. [19:01] I think I had the best luck with simulated annealing [19:01] :) yeah, it is a nice method. [19:02] hope I got it right... though all the methods I implemented should, if at all, contribute to the overall improvement in long optimization runs [19:03] the temperature schedule in CorewarOptimizer is set based on the number of iterations you input, so you need to wait for things to "cool" a little bit [19:04] in the next versions I'll probably also integrate an evolver and maybe some trained strategy classifiers once the syntax accuracy gets high enough [19:04] then the evolver could get the actual diversity of phenotypes and use THAT to alter mutation rates [19:05] instead of some predefined mechanism [19:05] but that's some far future :) couple of months from now or so [19:06] first need to add a bunch of other parametes, inputs... [19:06] maybe someone could actually code a "tree" -like structure on a evolver, like the mine that never got up from the ground [19:06] I've had good luck with randomized mutation rates.. [19:06] randomized? [19:07] yeah, I just set a range between pretty high, and low [19:07] Miz - maybe - it's a known methodology [19:07] and it randomly changes the mutation rate for each generation.. [19:07] @bw I also have a GA in the optimizer :) [19:07] and mutation in the rest of them, so I've toyed a bit... [19:07] seems to keep things from getting stale, but also keeps things from getting overly simplified by having a mutation rate thats too high. [19:08] Nenad: ah, i mustve spilled my guts on it some time ago in frustration, never mind then. [19:09] for instance, I've modified the hill climbing so that the probability of making large leaps = increasing the effective neighbourhood of an instance increases with the time from the last *good* score [19:09] to make it become able to escape low local extrema and such [19:12] GAs are nice, of course :) but one has to be careful with it [19:12] you know about NFL (!= National Football League) [19:12] ? [19:13] hey, thats a nice idea nenad. [19:15] well, NFL = No Free Lunch theorem basically states that when you take average performance of GAs over the set of all problems, they are no better than random [19:15] so you always need a good problem-tailored algorithm [19:15] just using some default is not recommended [19:16] anyway, I'm off to grab a meal now. [19:16] @bw - will you be online here these next few days? [19:17] me? [19:17] yup [19:18] yeah, I should be.. [19:18] ok, then we'll talk later [19:18] * Nenad waves [19:18] I'm never far from my email (unless I'm in dublin, where they don't let you leech electrons, and have very little wireless) [19:18] MSG: [21:37] MSG: Read error: Operation timed out [21:42] Join: johnkw joined #corewars [22:46] MSG: Quit: Lost terminal