[08:22] Join: Nenad joined #corewars [08:24] hi N [08:24] well.. the diversity hill hack does work.. [08:25] and it *DOES* keep diversity in the group [08:25] it slows down the forward progress quite a bit however.. [08:27] mostly because there's less to interbreed with successfully I think. [08:27] going to try tuning that a little, because the evolver was written to favour breeding pretty heavily over mutation. [08:30] hi @all [08:30] well, nice to hear that (evolver) [08:30] do you limit breeding to similar types? [08:30] or just according to scores? [08:31] how big is your population at some particular point in time - order of magnitude= ? [08:32] these were small test runs. so 50-100 warriors [08:33] and no, the breeding isn't really limited [08:33] I had tried that but it didn't seem to improve things. [08:33] well, then... [08:33] how? [08:33] how did you try it? [08:34] I tried by doing a "diff" of the warriors.. [08:34] according to what? - the instruction frequencies [08:34] ? [08:34] I'd pick a first parent, and then make a diff between it and the rest of the pool, and then pick from the top 25% [08:34] diff in scores? [08:34] no, diff in the code [08:34] so, the instruction frequencies? [08:35] 2 points for every insn that was the same, 1 point for every modifier, 0.5 points for every oprand [08:35] well, 'cause I think that true score-based clustering would go really really fast on a dataset that small [08:35] it may be worth the time [08:35] and a bunch of variations :) [08:35] you could identify true types in the population [08:36] well, score based clustering doesn't always do that tho [08:36] heh. it fails if the scores are really bad [08:36] because it varies with the opponents [08:36] yes? [08:36] oh [08:36] I forgot [08:37] so two warriors that are nothing alike, but significantly better than everything else end up breeding [08:37] you don't use the bench for the primary testing? [08:37] * Nenad is senile [08:37] no, this evolver evolves against itself, and there's a valhalla hill with the collected best of the best used as a bias group and a benchmark to let me know where the quality sits [08:38] so they battle themselves, but the scores also include a small random subset of the valhalla hill, to bias the pool towards human warriors [08:38] and then periodically, the best is tested against valhalla [08:42] how many people that are partly active here have evolvers? [08:42] * bvowk raises his hand [08:44] i am supposed to be coding the one.. [08:49] hmm... [08:50] I just meant to suggest that perhaps some sort of an evolving challenge might not be a bad idea, 'cause there hasn't been one for a long time now [08:50] but if there aren't people around to do it [08:50] :( [08:59] well.. drop a note on the r.g.c and maybe some people will show. [08:59] otherwise I'll claim victory without having to try.. ;) [09:02] that's a good idea actually.. [09:03] I'm planning to meet will for beers here later.. [09:03] so we could make a buy's the drinks thing ;) [09:11] bvowk: can you give a sample of two warriors that score quite differently according to your diversity points? [09:15] (and i mean a live sample) [09:18] two warriors that score differently? [09:18] or you want two that score the same but aren't? [09:19] i am mostly interested in what directions the warriors start moving towards if they dont have human bench to fight against [09:20] well, you dont use score-based clustering, so the question might be void [09:20] (ive been up for about 25 hours now..) [09:21] you want the code? :) [09:22] warrior's, not the evolvers [09:22] no, but if you had the code you could watch [09:23] i think i might try to implement that i-beat-the-one-that-beat-you based competitive diversity tomorrow, when i am more sober [09:23] they usually start moving towards mov towers, and they develop djn's or jmp-1's [09:23] you have the software logging behind https or similis? [09:24] I don't understand.. [09:24] jsut ignore my ramblings [09:24] heh [09:24] i need 5 more minutes before the caffeine kicks in [09:24] maybe you should try sleep.. ;) [09:24] it does a body good.. has what hackers need.. that sort of thing. [09:25] and i think ill go fetch my wonderbox of various perversions straight from Estonia [09:25] * Nenad says: who needs to sleep anyway [09:25] (dvd-r's and cd-r's) [09:25] bbl [09:28] mizcu: that may have been more information than required! [10:12] back [10:21] yay, 8cm dvd-r's [10:40] * Nenad observes the rain throw the window [10:48] *though [10:49] *through [10:49] omfg, I really need to learn how to type [10:49] * Nenad laughs [10:50] anyway, I'm a bit busy with other things atm, but - as for the warrior-evolving challenge - maybe you could come up with something interesting, bvowk? [10:51] standard or tiny or nano isn't that exotic. maybe something more unusual? [10:55] yeah.. I was thinking maybe an evolver writing contest :) [10:57] hrm.. [10:58] there are quite a few problems with that idea now that I think about it tho [11:00] maybe some constraints? [11:01] and a nice deadline - end of may or something like that [12:46] bleh. [13:08] what is bleh :) [13:08] it's like meh [13:08] no, not that. I know what it mean, damn it :) [13:08] *means [13:09] what I was trying to say was - why "bleh"? :P [13:11] its a feeling when you have somewhat uncomfortable feeling - tension in back or neck - and you describe you discomfort by saying 'bleh' [13:15] I have discomfort in my brain.. it's time to go home and stop working. [13:15] Est time: 248 days 14 hours 34 min 45 sec - 8177424 tiles remain [13:15] it's not going well today ;) [13:35] Join: crem joined #corewars [13:38] MSG: Ping timeout: 245 seconds [13:38] lol [13:38] well... [13:40] I'm a bit bored today, as well. Not in a mood to do much work, the weather is shi*** and I've spent most of the day searching for some papers, algorithms, reading math... so I feel like going to sleep right now [13:40] but that's not gonna happen [13:40] Join: crem___ joined #corewars [13:41] and my birthday is tomorrow :) but I'll spend the day at work. (sigh) no partying 'til weekend [13:43] MSG: Ping timeout: 245 seconds [13:44] Join: crem joined #corewars [13:45] MSG: Ping timeout: 245 seconds [14:04] happy bday nenad. [14:35] don't say that in advance :) [14:36] I'm not old. yet. :P [14:36] lol. kidding. well, thx [15:10] Join: crem___ joined #corewars [15:13] MSG: Ping timeout: 245 seconds [16:04] MSG: [23:10] MSG: Read error: Connection reset by peer [23:10] Join: crem joined #corewars [23:23] Join: Metcalf joined #corewars [23:36] MSG: Ping timeout: 245 seconds [23:37] Join: Metcalf joined #corewars