Code center > Suggestions
Catch-up mutation
spike43884:
Could you throw in a checkbox in global settings for Catch-up mutation which multiplies an individual bots mutation rates to 'catch up' to the average mutation in either:
(a) the rest of the species
(b) the rest of the simulation
It should help allow (fairly) controlled mutation for creating new bots more quickly as none get behind in slower mutation, but it won't overmutate them as when they catch up their mutation rate drops to normal again.
Botsareus:
Meh, I actually like my handycapped evolution hacks just fine.
If you think you can come up with better, hack at the source code like me for 10 years and we will see.
spike43884:
--- Quote from: Botsareus on April 10, 2016, 04:42:08 PM ---Meh, I actually like my handycapped evolution hacks just fine.
If you think you can come up with better, hack at the source code like me for 10 years and we will see.
--- End quote ---
Yeah but thats 10 years hacking at source code when for the next 8 to 10 years I need to be studying architecture....
Anyway, was it 10 years hacking at the source code, or was it 10 years performing zerobot evolutions with the occasional peak at the source codes evolution algorithms to evolve zerobots again? :P
Botsareus:
Actually I can see why you may request such feature based on how delta for zerobot mutation works. But trust me that with your idea it will be even worse. All the zerobots will end up evolving and dying without a chance for natural selection to kick in.
The zerobot algorithm was developed and refined for about only 1 year and I stopped work on it as soon as I got a result. I did not try it again after that. So I have no 100% proof that it even works consistently. I am going to revisit this soon.
The most time consuming algorithm is the survival evolution. When I first developed it, it was more effective than normal evolution. Now I am not sure if it still is mainly because of how much the program changed. However it still has clear advantages other than speed of evolution therefore in the long run my algo should be better. I am publishing this thing as a general algorithm to be used not only in db, so when it is prove of concept time, it should just work. Unfortunately prove of concept may take longer than earning my AA. One thing my algorithm did not change is how slow evolution is.
spike43884:
--- Quote from: Botsareus on April 18, 2016, 01:55:56 PM ---Actually I can see why you may request such feature based on how delta for zerobot mutation works. But trust me that with your idea it will be even worse. All the zerobots will end up evolving and dying without a chance for natural selection to kick in.
The zerobot algorithm was developed and refined for about only 1 year and I stopped work on it as soon as I got a result. I did not try it again after that. So I have no 100% proof that it even works consistently. I am going to revisit this soon.
The most time consuming algorithm is the survival evolution. When I first developed it, it was more effective than normal evolution. Now I am not sure if it still is mainly because of how much the program changed. However it still has clear advantages other than speed of evolution therefore in the long run my algo should be better. I am publishing this thing as a general algorithm to be used not only in db, so when it is prove of concept time, it should just work. Unfortunately prove of concept may take longer than earning my AA. One thing my algorithm did not change is how slow evolution is.
--- End quote ---
What about mutation which is increasing in specific area's of the simulation?
If it was thin fragments of elevated mutation, unless the bots .fixpos'ed in those spots they'd evolve at a varying rate?
Navigation
[0] Message Index
[#] Next page
Go to full version