Bots and Simulations > Evolution and Internet Sharing Sims

Project: New Ancestor

<< < (2/3) > >>

Shen:
The most important thing is the amount of energy going in and out of the field, which is turn related to how good your bot is at feeding and burning up its energy. The only thing I can suggest is to experiment with the bot your going to use and try and get a good balance. You want to try and make the enviroment as harsh as possible while still letting your bots survive. Apply some common sense to your sets, veg constantly being repopulated obviously needs more energy, veg constantly maxed out needs less.

I'd suggest using KE mode or possibly Zero Momentum (havnt tried that at length yet, might fun). As large a field size as you run at a reasonably speed. Around ~5 nrg for vegs though adjust it to suit. Also the mutation rates as slow as you can stand, the higher numbers the better chance of bad mutations being killed off leaving decent ones, I use around 25,000 to 50,000.

Carlo:

--- Quote ---Also the mutation rates as slow as you can stand, the higher numbers the better chance of bad mutations being killed off leaving decent ones, I use around 25,000 to 50,000.
--- End quote ---

Just an idea for mutations. I'd suggest to put much lower values for things such as gene duplication and gene deletion, etc. This because there are much more variables in a dna than genes, so gene mutations should be more frequent. If you have on average a gene every 10 variables, you should give 10 times more chances to gene mutation to happen, so you should put a mutation number ten times smaller.

Another idea, is to set the mutation rate of mutation rates to some very little value, say less than 10 (or even 1). A mutation of the mutation rates won't change the functionality of the dna (and, if I'm not wrong, won't be counted as a mutation). It will just mutate (sligthly) one of the mutation rates. This should eventually automatically balance the various mutation rates (and the mutrate of mutrates itself) to an optimal value for survival. This is because robots DO have interest in mutating: if they don't, other mutated robots will eventually become stronger and kill them or their progeny. So the point is to find the correct mutation rates, and this can be made through random mutations.

Numsgil:
Also don't be afraid to play with costs.  Setting store to a cost of 2 or 20 has some major implications for the bot's energy flow.

Shen:
Arnt the various mutation rates based on what there is to mutate? If theres more variables the bot will mutate more? So if you have 10 genes making up 100 variables both with 1 in 10,000 chance of mutating, you get a 1 in 1,000 for whole gene alterations and 1 in 100 for variable changes? At least thats what I assumed.

On the project note. Ive been messing with C_Ancestralis and decided it is possible to get it surviving in the new settings, its just hard. Might be fun to try and get it up to date purely through mutations. Are you interested in joining in Carlo, the more the merrier :)

Carlo:

--- Quote ---So if you have 10 genes making up 100 variables both with 1 in 10,000 chance of mutating, you get a 1 in 1,000 for whole gene alterations and 1 in 100 for variable changes?
--- End quote ---

Yes, it's like that. So, to have an appropriate number of gene duplications/ deletions, you have to set an appropriate value for gene mutations, that is, a value that takes in account the fact that genes are much less then variables.


--- Quote ---Might be fun to try and get it up to date purely through mutations. Are you interested in joining in Carlo, the more the merrier :)
--- End quote ---

Oh, thank you. But, you know, I run only ND simulations.  :lol:
Obviously, I'm joking. Tell me which version of DB to use and server name.

Navigation

[0] Message Index

[#] Next page

[*] Previous page