| |
| |
| |
|
Page: 1 2 3 4
Comments:
<0> oh man clips :'< I had to use that once
<1> <Magnus_RM@ef> I made a very simple AI chat bot, which some would not even call AI. But even though it is simple, it can make quite good chats. It produces sentences which are ready made in a way, or at least the structures of them, but with many alternatives in its building blocks (words or parts). So the result has quite some variety. The chat bot simply selects a random generator that has as many words as possible in common whith what you just s
<1> <Magnus_RM@ef> I also made a program that improvises music. Simple too.
<1> <Magnus_RM@ef> Both can be tested on the net.
<2> Magnus_RM: where can I test them?
<1> <Magnus_RM@ef> AIrelay, one second
<1> <Magnus_RM@ef> here are some samples of the music that my program "improvises" --> http://www.randomusic.com/samples/
<1> <Magnus_RM@ef> I will let my chat bot connect and put him in a channel so that you can test him
<1> <Magnus_RM@ef> AIrelay, did you see my replies to you?
<1> <Magnus_RM@ef> to test my chat bot Verbie, please join #verbie-chat
<3> Magnus: AIrelay is used to sync the #ai channels on multiple networks, the name given in <> is the person talking to you
<1> <Magnus_RM@ef> ops :)
<1> <Magnus_RM@ef> ok
<1> <Magnus_RM@ef> trane@fn, did you get my replies?
<2> Magnus_RM, yes
<1> <Magnus_RM@ef> ok :)
<1> <Magnus_RM@ef> so to test Verbie, join #verbie-chat
<1> <pacal@ef> what happens when you're already dehydrated and you drink beer
<1> <pacal@ef> does it just dehydrate you more or do you absorb some of the water in the beer
<4> how about another discussion night?
<5> moin
<6> moin
<5> cwenner did you get your books?
<4> no, didn't quite like those that were on that list
<5> ah ok
<4> i
<4> 'll look into it more tomorrow
<4> however, i found some articles on the subject :)
<4> http://portal.acm.org/ft_gateway.cfm?id=1194572&type=pdf&coll=GUIDE&dl=GUIDE&CFID=15151515&CFTOKEN=6184618 http://www.isys.ucl.ac.be/staff/luh/publi/achbany2005a.pdf
<5> nice
<5> man i cant await to get my new box :)
<5> i chose a bit bigger case now though with more space inside and bigger fans 80mm cause they are more silent so that if the fans are bad i can exchange them with super low noise fans if needed
<5> thats the case
<5> http://www.silverstonetek.com/products-sg01Evolution.htm
<5> this manufacturer is really cool
<5> and this is the psu
<5> http://www.silverstonetek.com/products-ST30nf.htm
<4> Did you figure out how to calculate the x displacement on the multivariate gaussians given the distance along the curve?
<4> bourbaki
<5> x displacement?
<5> ah
<5> well yes and no
<5> there is just a numerical solution
<4> ah okay
<5> if you consider this a solution yes i do not really
<4> that taylor expansion was sure interesting :p
<5> nah not really especially cause i wanted to be able to have a measurement for a lot of space
<4> i got to 11 or something on the train before i had to find a calculator
<5> taylor just good for close stuff
<4> hm, can't you approximate it as flat outside a certain range?
<5> i can but then it would be completely useless
<4> perhaps this range is then near enough to produce a goo approximation for the parts inside
<5> if i would continue my stuff id give the player a rubber band at hand
<5> that is its for a stragety game to measure how far you can walk
<4> well, wasn't it just a number of superimposed "3"-dimensional normal distributions?
<5> then youd get this band like irl and place it on the table top
<5> yes
<4> i don't know how inaccurate the taylor expansion was but if it's good enough near the mean (what's the tech. term in this case?) then ***uming it's flat too far off wouldn't be too bad of an approximation
<5> hm ... close by ?:)
<5> near the mean is fine i think
<5> its a lot easier to just use this band thingy
<5> do you know warhammer?
<5> the table top game
<4> i mean if the taylor goes hiwire at a distance 20 but the variance is just 5, then it's okay to presume that everything beyond 15 for instance is flat
<7> can anyone here suggest some good software for image processing?
<5> i tried to approximate the distribution with taylor but even this didnt work out properly
<5> crm_ opencv
<4> how far was it within 10%?
<5> http://en.wikipedia.org/wiki/OpenCV
<5> cwenner it was ok if you were close to the middle
<4> how close?
<5> but since i needed a function that also works on inf it didnt worked out at all
<4> if it's for instance up to 1 variance, then you could surely get it good *enough* by expanding it further
<5> something around 0.6 id guess
<4> well at distance inf, you can ***ume that it's flat :D
<5> square dist
<5> thats what i do now
<4> so why not try and expand it 4 times further
<5> i build my own polynomial by hand though which is smaller than the taylor series
<4> and see if you can get it up a bit more
<4> and ***ume that everything outside the range is flat
<5> i tried it with like 8
<4> it seems feasible
<4> to me
<5> but the series was worse than the polynomial that i build
<7> its a windows library Bourbaki ?
<4> built your own?
<5> crigo yes and linux
<4> hmm, really
<5> cwenner yes i used a cubic spline and scrapped some stuff
<5> now its like 4 adds and 3 muls
<4> still seems to me that it's doable
<5> quite fast
<5> taylor would need a lot more
<4> if it's this gooda lready, then just getting it a bit better allows you to get an upper bound on the error for all values
<5> plus i used the tayolr to approx exp
<4> yes, you told me
<7> nice.. thanks Bourbaki
<5> np
<5> did you see the last video cwenner?
<5> of my app that is
<4> havent seen any video, bourbaki
<5> ah sec
<5> http://video.google.de/videoplay?docid=-979203120692178262&q=fluid+solver
<5> thats the stuff in action with my fluid solver
<4> perhaps just my imagination but the first video seems to be a lot better than the otehr two
<4> not sure what it manages yet
<4> a more bumpy environment would be nec for that
<5> yes its different setups
<5> the first video has no adherence
<4> can it handle step hills for instance, such as /----\
<5> so the particles dont attract each other
<5> well if you can build it with these hills yes
<4> they didn't seem to move as smoothly either
<5> its like FFT
<4> that's what i was implying
<5> just another base function
<4> interesting to see how your engine handles such a, possibly harder case
<5> i thought about adding other functions also though
<4> where the hills are step in comparison to the particles
<5> not that hard but id need it for games like populous
<4> how do you handle caves?
<5> hard edges wont work cause you cant derive them
<4> neither two planes
<4> hmm.....
<5> caves might be possible if you combine an upper and lower limit
<5> that is two functions one on top that models the ceiling
<7> do many regulars use opencv Bourbaki ?
<5> and one that modles the bottom
<5> crm_ do you mean regulars here?
<4> would be interesting to see, bourbaki
<7> Bourbaki yea
<5> opencv is widely used but i dont know if its used by the guys in here
<4> perhaps if you model gu***ians on spheres
<5> i think theres also a computer vision channel on the server here
<7> you know what its called by any chance?
<5> cwenner i can do that
<4> #machinevision, not very active though
<5> the mapping is crappy though yet
<5> but i can do it on tori really nice
<4> i meant the outside of a sphere but pretty much the same thing
<4> oh
<5> crm_ the lib is used by a lot of ppl though so youll find good support on forums and such
<5> cwenner i have a really good caculus under all of this
<5> you can map the gradient field into any other derivable 2 manifold
<5> for a sphere its not that nice yet cause a sphere has another topology than a plane
<5> so for a sphere id need to adapt the calculus a bit to make it work 100%
<5> its looks a bit arkward at the seams
<5> in the video you can see though that the simulation runs on a torus
<5> its just not mapped to a torus cause its hard to stir on a torus
<4> derivable 2 manifold = a manifold that is derivable twice?
Return to
#ai
or
Go to some related
logs:
opensuse unichrome xgl
error while loading shared libraries: libpcre.so
yum DB_RECOVERY
dillo Gtk-WARNING **: gtk_widget_size_allocate()
howto autostart mdadm in gentoo
wperl vs perl
#xine
#math
do_stage ()
+javascript +removeChild +an object exists
|
|