[a / b / c / d / e / f / g / gif / h / hr / k / m / o / p / r / s / t / u / v / vg / vr / w / wg] [i / ic] [r9k] [s4s] [vip] [cm / hm / lgbt / y] [3 / aco / adv / an / asp / bant / biz / cgl / ck / co / diy / fa / fit / gd / hc / his / int / jp / lit / mlp / mu / n / news / out / po / pol / qst / sci / soc / sp / tg / toy / trv / tv / vp / wsg / wsr / x] [Settings] [Home]
Settings Home
/g/ - Technology

Thread archived.
You cannot reply anymore.

File: 1451264353269.png (629 KB, 1000x1000)
629 KB
629 KB PNG
>endless talk of genetic evolution of computer programs
>no one ever talks about how those genes translate to other data formats
What are you even talking about? Do you want me to do? Program a genetic learning AI and save the text file as an jpg?
I'm talking about how no one explains how a genetic sequence becomes a computer program, in such a way that it can breed with similar genetic sequences to produce a mutation that is less likely to be awful.
Did you skip your physics classes?
Genetic algorithms try to tag results with a number/score i.e. good or bad.

They will evolve with the aim of improving that score, it depends on the context but it'll produce images or whatever format according to the rules you've set; in order to improve that score.
File: Gospers_glider_gun.gif (21 KB, 250x180)
21 KB
Even with a small handful of simple rules, you can wind up with a self-replicating system that grows in complexity. Just look at Conway's Game of Life.
It has literally nothing to do with physics.
It's pure computer science.

I don't think you understand what I'm talking about.
Implementing natural selection is absolutely trivial, but the thing that no one explains about genetic algorithms is how a test is made from a set of genes.
you can easily represent the data as as set of bits, you can easily apply a mutation operation and in that case, your chromosome is that chain of bits:

000011 (mutation)

111000 (splicing)

As long as you can make provide a mutation and a splicing operation, you can mutate anything. Does that help?

I took that class ages ago but it's mostly what I recall
>you can easily represent the data as a set of bits
Yes, but it's not always easy to do it in such a way that with two sets of bits, every bit is analogous to the other bit belonging to the same address; and that's important for crossbreeding.

Although I guess it's only a problem if you have something like, two neural nets with differing numbers of neurons.
That's the catch, you have to be able to logically represent it as something you can mutate.

If you had floats for instance, a mutation might be +- 0.1f instead of plain bitwise operations, as the process of evolution is typically 1.2f -> 1.3f and not 1.2f -> 231421f
do you know what a fitness function is, or any idea of statistics whatsoever?

Delete Post: [File Only] Style:
[Disable Mobile View / Use Desktop Site]

[Enable Mobile View / Use Mobile Site]

All trademarks and copyrights on this page are owned by their respective parties. Images uploaded are the responsibility of the Poster. Comments are owned by the Poster.