General > Biology
Neural networks using bots
Numsgil:
That's sorta the area that hasn't been wikified yet (yes, it's a verb now). To tell you the truth I've never played with them so I'm not too sure either. I think PY's multibot tutorial might have something about them, or even the general help file in game.
Ulciscor:
OK [Num] thanks I think I'll mess about with them now and see what I can find out!
Ulciscor:
OK well I have an idea for the energy sharing network but I don't know if it is overcomplicated or if it will even work.
I was thinking every time a new bot is born it bonds to the nearest 4 relatives and generates a random number for itself, from a suitable large range. So each bot in the net would hopefully have a unique ID.
So when a bot's energy is below a certain value, it would send it's ID to the attached relatives. They would then send out the ID to their attached relatives. Then when a bot receives an ID that is not the same as its own it would send some energy to all attached relatives. When a bot received energy and it hasn't got the same ID as the one it received it would send the energy out on all connections again.
Now I read what I have written I've just realised the whole thing could get stuck in a loop with the same bots sending each other energy all the time without it getting to the right bot.
Help!!
Light:
sounds like you want .sharenrg
Numsgil:
Yes, but that would defeat the purpose to some extent.
I think you're on to something Ulc, you just have to apply a damping force to prevent that kind of infinite oscilation.
In a spring the force is -kx - vb where k and b are constants. The -vb is what stops infinite oscillations. In your case, x is energy requested - energy wanted and v is energy just sent last cycle (*.pain). (I might have some signs backwards, bear with me). Keep sending energy out/collecting energy at rate v every cycle, where v changes by -kx - vb as I described above.
If that didn't make any sense I can try again. I'm a little tired.
Navigation
[0] Message Index
[#] Next page
[*] Previous page
Go to full version