[00:05] Join: flyduck joined #corewars [00:05] halo [00:06] halo [00:08] hmm, let's see how well pycorewars works [00:08] Hi Flyduck [00:09] * AndrewBC sits on flyduck [00:29] whats up guys :P [00:29] working on anything new ? [00:30] * AndrewBC is setting up for evolution [00:30] hand-coding is frustrating me a little bit, because I've -almost- gotten on the nano board a few times now [00:31] so I will employ some mutation and evolution on my already written bots, as well as some randomized ones, and pit them against bvowk's metric ton. [00:35] bvowk, your page makes me giggle [00:39] Hmmm, PyCorewar is looking pretty decent, judging by it's examples [00:44] AndrewBC, are you writing the evolver yourself:? [00:44] i dont know what PyCorewar is, heh :P [00:44] Yes, I'll be writing it myself, and if it turns out to actually work well, I'll release it under the MIT license [00:45] http://corewars.jgutzeit.de/pycorewar/index.en.html [00:45] pretty much all I need to be able to write the evolver in my most comfortable language [00:45] i started writing an evolver myself, im about 80% done with it [00:46] yeah? :) [00:46] i just got a little tired of it and went back to hand-writing warriors about a week ago :P [00:46] heh [00:46] but ill pick it up again [00:46] indeed [00:46] im toyed with genetic algorithms before, i wrote a program that evolves a rational approximation of PI to learn :P [00:47] ive * [00:47] * AndrewBC 's eyes go glassy for a moment [00:47] ... what? I have no recollection of the past 20 seconds [00:47] you know, like a fraction that approximates PI :P [00:47] to learn? [00:47] yeah ! :D [00:47] to learn what? [00:48] and how? [00:48] genetic programming ! [00:48] hehe [00:48] * AndrewBC 's eyes go glassy for a moment [00:48] :P [00:48] i can send it to you if you wanna see, its in perl :P [00:48] the best it has given me is PI to 6 decimal places with the 7th being off by 2, heh [00:48] nah, thanks though [00:48] I'm too occupied as it is [00:48] with being sick >:( [00:48] PyCorewar looks nice ! [00:48] too bad im not a python programmer :x [00:49] flyduck, oh, it's learning to approximate py via genetic programming [00:49] (i dont know if i should start reprogramming my current evolver to what i planned, it gave me enough headaches as is) [00:49] gotcha [00:49] er [00:49] pi [00:49] .... hah, whoops [00:49] AndrewBC, exactly, heh [00:49] Mizcu, what about it gave you headaches, and why? [00:50] the most annoying part for me of a genetic alg for a warrior is the fitness function... what do you pit it against ! [00:50] i had some functions that i couldnt understand why they didnt work [00:50] flyduck, for some reason I thought it was somehow approximating pi in order to learn something else [00:50] first problem was too obvious to solve at first look, second problem was.. i still have no clue [00:50] nah, its a really basic algorithm, it was like a 'hello world' for genetic programming, i figured [00:50] i was surprised and delighted when it actually worked :) [00:50] Mizcu, well if you ever need a second set of eyes [00:50] flyduck, nice :) [00:51] AndrewBC: a rnd function gave different results when being called from different functions, is that creepy enough? [00:51] not yet, I have no idea what an rnd function is [00:51] function that returns a random number [00:52] oh, hm [00:52] different results? [00:52] isn't that the point of it? [00:52] what do you mean by different results? [00:52] it gave results out of the limits given to it [00:53] was supposed to return between 0 and 12, returned massively wrong numbers [00:53] is this in Visual Basic, or what? [00:53] C [00:53] oh [00:54] i did actually switch to another rnd-function used in the software (i manipulated existing software) and it started giving reasonable numbers again [00:54] yeah, maybe that one was just bad, or something [00:54] there was no datatype conflict either, which i wouldve considered the most logical reason for a failure [00:56] crazy stuff [01:26] ooh, I just had a good idea for a pseudo-scanner that would require less room [01:26] evolving will have to wait a bit :X [01:29] hehe :D [02:12] MSG: Quit: alexeld [02:45] if i use # as the modifier for a SEQ offset field, will it compare to instruction in the other field to the number 0 ? [02:48] remember that by default it's seq.b, so if the other field is not immediate, it will be looking at the b field of wherever it pointsd [02:49] so if you did "seq #0, 0" that would work, because the b field of the relative instruction 0 away from that line is 0 [02:50] i though seq #nnn, $nnn defaulted to seq.ab? [02:52] does it? [02:52] it would take almost 15 seconds for me to test [02:52] me too! [02:53] it does [02:53] my mistake, flyduck [02:53] what other commands don't default to .B ? [02:53] I thought they all did [02:54] defaulting depends on the addressing modes used [02:54] mov $nn, $nn will default to .i [02:54] that makes sense [02:56] * AndrewBC crosses fingers [02:57] oh that was awful [02:57] just awful [03:04] MSG: Quit: Leaving [03:11] new warrior ready [03:12] ^~^ [03:21] lets see how it fares.. [03:27] 69.3 :/ [03:28] bout as well as previous versions [03:55] :( [03:59] i think i need to read more at this point [03:59] im making better warriors but my understanding of strategy / optimization is limited... [04:00] yeah [04:00] same here [07:14] MSG: Quit: flyduck [09:19] MSG: Quit: This computer has gone to sleep [11:10] Join: sh0ne joined #corewars [11:10] Hi [11:11] Join: fish joined #corewars [13:07] back [14:41] front [16:58] MSG: Quit: z/ [18:06] MSG: Remote host closed the connection [18:21] Join: baudchan joined #corewars [18:54] Join: sh0ne joined #corewars [18:54] Hello [19:08] MSG: Quit: This computer has gone to sleep [20:32] Join: flyduck joined #corewars [20:44] Heya sh0ne [20:48] Join: baudchan joined #corewars [20:57] MSG: Read error: Connection reset by peer [21:01] Join: sh0ne joined #corewars [22:45] MSG: Quit: Leaving [23:06] Join: Altair joined #corewars [23:37] MSG: