12 / 1

4th April

10th April

the basic foundaition that modern neural networks are built on today, biased randomess. this proof of concept can't lean and has very limited capabilitys but is cool nonetheless. go to ID:3360593 to see this in practice.

artifiacial intellegence human circuit neural network computer

Comments

DRZOIDBERG 10th April

klao003: especially theese, since they are just a demonstrition of a concept
klaotte003 10th April

NotAChair: those are way too simple for that
DRZOIDBERG 10th April

that would be cool but as they have roots in randomness it may br hard to be consitant without having to do it multiple times. or maybe we jsut need to get better at making them instead lol.
NotAChair 9th April

I think we should hold a contest to see who can make a learning neural net that can operate a simple nuclear reactor(or other task with relatively few inputs and outputs).
ThingamabobOverseer 9th April

In the beninging
five_pebbles 8th April

absolute begginigs +1
DRZOIDBERG 6th April

I mean, it's output can mimick that, but the outputs can have several areas of high frequency and it all depends on where you chose for the biases to be. the central bias does produce a graph that is similar to the gaussian distribution but that is only due to it only hhave one central area of bias surrounded by equal lower biases.
klaotte003 6th April

so it uses gaussian distribution technically
DRZOIDBERG 6th April

the one-way junctions are the main way and the wifi is mostly just there to make a feedback loop
DRZOIDBERG 6th April

the bias can be controlled with wifi placement but also be junctions between neurons that only travel one way

The Powder Toy

the begginigs of a neural network by DRZOIDBERG

Comments