| |
| |
| |
|
Page: 1 2 3 4 5 6 7 8 9 10 11 12
Comments:
<0> lol
<1> where that ip is registered.
<1> in any case.
<0> he is from canada dang it!
<0> blame canada!
<1> expert systems aren't like 'hello world'
<2> hahah yeah
<1> it might be beyond the scope of this irc channel to teach someone how to make one.
<2> seriously.. i need help with expert systems too, for that matter
<1> me too :P
<0> goodie
<2> haha
<1> i'm working on a neural net for image processing.
<1> which will eventually contribute to the neural net that'll be doing image analysis.
<2> wow for what application?
<2> we could use something like that in the neuroscience research lab i work in
<1> heh
<1> probably not.
<1> the first one, the easy one...
<1> i want to compare my idea against bilinear and bicubic interpolation.
<2> i am not familiar with those algorithms/techniques.. what do they accomplish?
<1> i figure neural network generalization properties could prove to be great for scaling up images.
<1> they're used to scale up images.
<3> "[11:07] <1> expert systems aren't like 'hello world'"
<3> True.
<3> std::string name; std::cin >> name; std::cout << "Hello, " << name;
<3> THat
<3> That's an expert system.
<1> where's the inference engine?
<3> Oh, it's supposed to figure out who's typing?
<1> no, it's supposed to have an inference engine.
<1> "A software system with two basic components: a knowledge base and an inference engine. The system mimics an expert's reasoning process." - google
<1> you could argue that cin>>name establishes a knowledge base..
<1> but cout<<name doesn't qualify as an inference engine.
<3> std::string name; std::cin >> name; std::cout << InferStuff(name);
<3> Duh
<1> heh
<1> no :P
<2> an inference engine is one of those thingies that has a bunch of implications in it and can infer a bunch of conclusions given some truth statements, right?
<2> sort of how prolog works?
<1> basically.
<1> there's numerous ways of implementing an inference engine
<2> man this reminds me how much i miss computer science and how nice it would be to go back to school..
<1> i'm a fan of neural networks myself, they have some cool properties that you won't find in a heirarchical-conditional tree.
<1> namely the ability to generalize a solution for inputs that weren't in the original training data.
<1> and the ability to learn by being trained with a data set, thus avoiding the need to explicitly determine a solution to the problem.
<2> wow
<2> i am reading the theory behind them now
<2> i am actually shamefully infamiliar with how that works in detail :(
<1> so was i until relatively recently :P
<1> it's not as complicated as it seems.
<1> well, the basics at least.
<1> and then you start building your own network and you're like "wait, how do i know how to interconnect my neurons..."
<1> and yea :P
<2> i am fascinated by this
<1> so basically what i'm working on now is building a neural net that can learn a graphical image...
<1> you train it with the pixel values at each integer pixel..
<2> i still don't fully grasp it.. i need time to really think about the implications of this simple thing i am reading about
<4> we're using it for gyroscope-based extension range measurement in arthritis patients
<1> 0,0; 0,1; 0,2; ...; 1,0; 1,1; 1,2; ... etc.
<1> and then after it's sufficiently trained, you ask it for pixel values at a higher resolution.
<1> 0,0; 0,0.1; 0,0.2; ...
<1> effectively scaling up the image 10x
<1> and then i'll compare it again popular up-scaling algorithms, like nearest neighbor, biliear, and bicubic interpolation.
<2> sounds simple in concept.. but i still don't really get how you would implement something like this..
<1> and also against a perfect up-scaling (i'll start with a large image and down-scale it to create my 'source' image for the study, but still retain the original large one as a 'perfect' up-scaled version)
<2> ah i see
<2> rigght
<2> that sounds cool.. are you a phd student?
<2> or doing this for fun?
<4> neurons are pretty nicely implemented with OOP
<1> heh
<1> i'm a senior in my undergrad right now
<2> oh hah
<1> and yea, i used c++ oop for the implementation.
<1> although i'd have to find a much more efficient way of doing it were it to ever be used in commercial graphical image editing software :P
<1> either way, i'm thinking about using this as my masters project
<1> or writing it up for ieee or something.
<1> or both :P
<2> wow nice
<2> that is impressive
<2> i am surprised this hasn't been tried yet
<1> i wanna get published, that's the only way you can really gain credibility.
<1> yea, so am i.
<2> have you looked into the existing literature?
<1> i have.
<1> well
<1> some address the basic idea..
<1> especially one that just came out that i haven't seen yet :\
<1> wtf
<0> lol
<1> japanese bastards! no!
<2> what?! they did it?
<1> "Image Interpolation using Feedforward Neural Network"
<1> great. thanks guys :\
<1> looks like i at least had a publishable idea.
<2> well what was their conclusion
<1> "The experimental results are shown and evaluated. The effectiveness of our methods is discussed by comparing with the conventional methods."
<1> i don't know, i'm not about to buy the paper.
<5> Your university will probably have it
<2> how much could it be?
<2> yeah also if you are at a big uni you probably can find the paper, like Asmodee`
<2> wem Asriel said
<2> ok this convinces me i need to go back to school and into some phd program.. i find neural nets fascinating
<1> i'm not at school right now. it's sunday.
<1> or monday.
<1> one of those :P
<5> you don't have remote access
<5> ?
<1> they offer vpn access, which i'm not interested in.
<6> 'morning
<1> hiya run.
<1> i have an ieee membership too
<5> morning run
<2> hey ninevolt can you tell me where that paper was published and/or by whom? i want to read it too when i get to work..
<1> but they restrict your access to papers to only ones in the areas you subscribe to.
<1> http://www.google.com/search?q=Image+Interpolation+using+Feedforward+Neural+Network
<1> top 2 scholarly articles look relevant, as does the first www hit.
<2> wow i never knew google did this scholarly articles thing.. hot damn
<1> yea, it's nice :P
<2> google is so awesome
<1> yea, my 'big' idea might make google images more awesome.
<1> if google has abundant computational power.
<0> heh
<0> you cant
<6> Can some function declaration contain a quote?
<2> i don't get run's question
<6> You can, note the 'possibly'
<7> how would you do that Run
<0> there are instances where they are the same
<1> Run is carlo@127.0.0.1 * Captain Picard
<1> o_O
<2> am i crazy? what is Run asking? and how is it you understand him?
<1> run knows more than i.
<6> There are two types of questions that you can't understand.
<1> he's irc'ing as localhost :P
<6> 1) The ones that make no sense whatsoever.
<2> i don't doubt run's intelligence, i am merely wondering what he asked?
<6> 2) The ones that you are too stupid^H^H^H^H lack the experience for to understand.
<2> haha
<2> i never claimed your question fell into category (1)
<6> Lemme give an example...
<6> need somet time for that...(need to write code)
<2> no need to get polemic here.. can you just clarify?
<6> well
<6> C++:
<6> void Foo::Foo();
<6> That aint C.
<6> Possibly C:
<6> void Foo();
Return to
#c++
or
Go to some related
logs:
ipp2p hamachi
tattoo paceville
#AllNiteCafe
#linux
#linux
#linux
#AllNiteCafe
#linuxhelp
#AllNiteCafe
bcb imagemagick
|
|