Author Topic: Only delta mutations! RESOLVED 2.43.1h  (Read 3397 times)

Offline Testlund

  • Bot God
  • *****
  • Posts: 1574
    • View Profile
Only delta mutations! RESOLVED 2.43.1h
« on: April 06, 2008, 12:06:49 AM »
Can somebody explain to me why I haven't got a single point mutation on any bot for over 300000 cycles? I have point mutations set to 3200. I used to get it a lot more often when I had it set to 5000. The smaller the number the more often the mutations should happen.  
« Last Edit: April 06, 2008, 03:57:00 PM by EricL »
The internet is corrupt and controlled by criminally minded people.

Offline Testlund

  • Bot God
  • *****
  • Posts: 1574
    • View Profile
Only delta mutations! RESOLVED 2.43.1h
« Reply #1 on: April 06, 2008, 03:15:55 PM »
I've done a lot of testing and tweaking on the mutation rates this evening to find out what's going on. Either I managed to circomvent a bug or Eric has done changes to how the mutation rates work.
These are the values I've typed into the mutation fields which I think gives the most decent mutation rates:

Point: 3

Copy Error: 36

Reversal: 432

Delta Mutation: 4000

Insertion: 144

Minor Deletion: 144

Major Deletion: 216

This was tested with 10 zerobots and 10 Alga Minimalis where I have set reproduction to occur when the bot reaches 1000 energy.

I have found that the lower I set Delta Mutation the less I get of other mutations. Setting Delta Mutation to something like 3000 will cause other mutations to become very rare. If I set Delta Mutations to a very low value, like 30 or something I will only get Delta Mutations in the sim!

Here's a sim if you want to check it out by fiddling with the mutations:
The internet is corrupt and controlled by criminally minded people.

Offline EricL

  • Administrator
  • Bot God
  • *****
  • Posts: 2266
    • View Profile
Only delta mutations! RESOLVED 2.43.1h
« Reply #2 on: April 06, 2008, 03:41:10 PM »
This is my fault.  I left a test line in that effectivly disabled point mutations.  Fixed in 2.43.1h, which should be out in about 5 minutes.
Many beers....

Offline Testlund

  • Bot God
  • *****
  • Posts: 1574
    • View Profile
Only delta mutations! RESOLVED 2.43.1h
« Reply #3 on: April 06, 2008, 09:47:55 PM »
Ok, I've done a lot of testing now with version 2.43.1h and now I need to increase point mutations to 640 to get a good balance. The other values still need to be pretty much as above. Also I seem to be able to have a lower value now on Delta Mutations without it overriding the other mutations. So something is still different from earlier versions way back, because such low values would just have destroyed the bots previously. I don't know if anybody else here get the mutations right if set to the default values of 5000.
The internet is corrupt and controlled by criminally minded people.

Offline Numsgil

  • Administrator
  • Bot God
  • *****
  • Posts: 7742
    • View Profile
Only delta mutations! RESOLVED 2.43.1h
« Reply #4 on: April 06, 2008, 09:59:30 PM »
The default of 1 in 5000 was quite literally a wild guess when I was setting up the mutation rates.  They aren't balanced regarding each other or even what a sane value might be.

Offline Testlund

  • Bot God
  • *****
  • Posts: 1574
    • View Profile
Only delta mutations! RESOLVED 2.43.1h
« Reply #5 on: April 07, 2008, 01:17:53 AM »
Well, I can't imagine how a programmer can make this work in the first place.
The way I think would be optimal is if you could make it so the program throws a dice for each bot, like if you have point mutations set to 5000, then it should be like throwing a 5000-sided dice every cycle for each bot. If 5000 comes up then a point mutation occurs. The same for the copy mutation, but instead it throws it everytime a bot multiply.
Maybe that would be too much calculations for the program, especially if you're running with thousands of bots.
The internet is corrupt and controlled by criminally minded people.

Offline Numsgil

  • Administrator
  • Bot God
  • *****
  • Posts: 7742
    • View Profile
Only delta mutations! RESOLVED 2.43.1h
« Reply #6 on: April 07, 2008, 01:25:32 AM »
It assumes that the probability of a mutation is a Poisson distribution.  Most mutations are per bp.  Based on the probability for a mutation, the program calculates when and where the next mutation will occur, assuming that mutation events are entirely independant events.  After a mutation is triggered, the program calculates a new when and where.  It's entirely possible using this system for two mutations to occur on the same cycle.