Neural Network

I don't get how neural networks work.  I've looked at wikipedia and some example code, but I still don't get it.  Can someone explain it simply?

A neural network is a simulation of a circuit where every node ("neuron") is connected to every other node. The strengths of the connections determine the behavior of the whole circuit. It's named after neurons because that's how the brain works on a very low level, not because it's a feasible building block for every part of an AI.

[spoiler]faggot[/spolier]

*grabs dick*

>>1
They don't, really. They're just a bunch of things called neurons, which each take an input value and produce an output based on it. A neuron can (and presumably does) feed into one or more other neurons. Usually they form a very simple structure, such as a few neurons taking input values, which are wired to a hidden layer, the outputs of which connect to output neurons. Making this do anything useful is complicated, since it's not possible in general to predict what changes to neuron weights or wiring will do. The more hidden layers you have, the harder it gets. The more complicated your wiring gets, the more complicated it gets.

If you can't understand the basic components of a neural network you're kind of dumb, but if you're just not sure what to do with them, rest assured you're not alone. Neural networks are dodgy things.

>>5
I'm not the OP here, but your post actually explained them to me without all this eleven-page wikipedia bullshite (which I can now read).

>>5
I use a simpler and faster model:weighted choice array, i'll post on it later on my blog.
Basically its an array of function pointers and their weight, the selected (randomly, higher weight of choice =more chance of selection) pointer performs chosen action and adjust its weights, based on response to action(e.g. func_weight*1.01).


_____________________________________
http://bayimg.com/image/aadbjaace.jpg
Velox Et Astrum gamedev forum: http://etastrum.phpbb3now.com
Behind every failure is an opportunity somebody wishes they had missed.

I use a better and better model:fattness choice array, i'll post on it later on my blag.
Basically its an array of macro pointers and their fattness, the selected (randomly, higher fat of choice =less chance of deletion) pointer performs chosen fuck and adjust its fat, based on response to trolling(e.g. func_fat*11d7).

Here maybe some ascii would help


o-o-o
 x x
o-o-o
 x x
o-o-o