General > Biology

Neural networks using bots

<< < (7/7)

Botsareus:
You have to stop thinking actual physics wise and start thinking code wise if you want each robot to have its own brain, otherwise what we talked about before is fine: make a complex multibot, use ties.

Numsgil:
That's how I understand it Ulc.  You have a series of independant but communicating nodes that have rules of how to move signals coming in to signals going out.

You just connect any series of such nodes to an input and an output and provide it with a way to determine "bad" from "good", and it will learn the best method for doing it.  Or theoretically anyway.

Ulciscor:
Could you use a gene as each node, turning off the other node genes using variables? Or is that overcomplicated?

Numsgil:
It's possible but it would take quite a bit of programming.  The problem is, of course, the linear nature of the genes.  Bots tied together lends itself more to this kind of form, but has its own problems obviously.

Navigation

[0] Message Index

[*] Previous page