Return Styles: Pseud0ch, Terminal, Valhalla, NES, Geocities, Blue Moon.

Pages: 1-

neural networks

Name: Anonymous 2018-10-27 19:48

redpill me on neural networks

what is a neural network? how do I make one? why would I want one? what can you use them for?

Name: Anonymous 2018-10-27 20:06

inb4 mentishit think it's the perfect occasion to spam his shitty random sentence generator.

Name: Anonymous 2018-10-27 20:28

>>2
Every occasion is perfect occasion to learn about Ghost Mind Algorithm.

Name: Anonymous 2018-10-28 13:54

I guess you could call it a reliable way to implement fuzzy logic
fuzzy in that while there isn't any simple rules for telling between a picture of a cat and a dog, with a good set of examples you get a fuzzy predictor of cat v dog pictures

Name: Anonymous 2018-10-28 16:08

>>4
would it say like "58% sure this is a dog" instead of "this is a dog"?

Name: Anonymous 2018-10-28 16:17

>>5
Did you just assume its species? Its could be 58% dog and 40% cat, with bear heritage, identifying as a bear.

Name: Anonymous 2018-10-28 16:29

>>6
but we're all just americans amirite

Name: Anonymous 2018-10-28 16:32

Basically, neural network is a composition of addition and multiplication operation, plus a special algorithm that finds proper multipliers, so that the composition begins to approximate some specific function. For example, you can train neural network to approximate sin or log functions.

"a*w1 + b*w2 + c*w3" is a neural network. The complexity arises when a, b and c are neural networks too. And you have to include several layers, because single layer can't approximate non-liner functions.

Name: Anonymous 2018-10-28 16:34

>>8
Oh, the training algorithm tries to pick proper w1, w2 and w3. Basically, the training algorithm is the hardest part of neural network theory, specially if neural network is self referencing.

Name: Anonymous 2018-10-28 18:52

neural networks will be proven to be a meme once its unpredictable nonlinearity starts running over and killing people

Name: Anonymous 2018-10-28 19:40

>>10
Neural networks are harmless, as long as they are not a part of self preserving system with some goals.

Name: Anonymous 2018-10-28 21:22

Name: Anonymous 2018-10-28 22:49

>>12
So women are the first real victims of AI?

Name: Anonymous 2018-10-29 2:18

>>5
58% is a bit misleading, since 50/50 is 0% sure of it being a dog
That would really be more like +8% or +16% depending on scale (-50 to +50) or (-100 to +100)

Name: Anonymous 2018-10-29 7:03

>>13
redirect male sexual desires
less rapes and molestations
rapists turn to sex robots, avoiding jail
feminists screech about it: WE NEED MORE RAPISTS

Name: Anonymous 2018-10-29 7:15

>>15
who are you quoting?

Name: Anonymous 2018-10-29 7:46

>>16
Its not my job to educate you, fascist.

Don't change these.
Name: Email:
Entire Thread Thread List