Darwinbots Forum
Bots and Simulations => Evolution and Internet Sharing Sims => Topic started by: Greven on June 27, 2005, 07:19:49 AM
-
Anyone interested to make a guide about analyzing evolved bots?
Maybe even writing a small program helping people out!?
If a evolved bots DNA has increased by a factor of 10, it is extremely diffecult to analyze it, like my own evolved violet bots.
Any comments, ideas etc.
-
We dont need a guide , we just need better tools.
-
My advice is not to let it go beyond recognition - save some intermediate bots and align those to each other. The system with unmutable tags that I proposed some time ago would make the whole process automatic, but it did not look like Nums liked it, so I doubt it will get implemented.
The tools (like debugging/monitoring tool I proposed a long time ago) would also be helpful. I think something is being done along these lines, so ask PY or Nums on that.
-
Saving the DNA from within the program leaves the text file as a almost unreadable list of numbers and sysvars. But if you open the bots DNA in the info window and copy it out its much better.
Other than that, try and memorise the basic sysvar numbers (7 is shoot etc) so you can work out some of the intergene stack fun.
-
My advice is not to let it go beyond recognition - save some intermediate bots and align those to each other. The system with unmutable tags that I proposed some time ago would make the whole process automatic, but it did not look like Nums liked it, so I doubt it will get implemented.
Don't you have that backwards shvarz? I thought I proposed the unmutable tags and you were the one who didn't like it.
That was a long time ago. Either way, I'll put the idea in my list.
-
Okay! Just a suggestion, nothing more. Thought people had some interesting ideas/experince with this one. It just seems impossible to decipher a DNA with over 4000 commands scattered in 12 genes! :wacko: Yes, if you really want it, you do it, but it is terrible time consuming, and just a little mistake may wreck the entire belived behavior of the bot! :(
Basically we cant see at the screen a bots behavior, a lot of bots do the same thing, but some are more efficient than others, and here small changes in the DNA can make a huge difference.
What about describing some of the common trait you all programmed that makes a bot better. Not nessacary genes, but rather the behavior of such thing? Like the non-spinning gene I the simple bot etc., hope you understand what I mean. I like to get a more biological view on DB and its bot and behaviors. And have anyone seen Alien Planet on Discovery?
-
And have anyone seen Alien Planet on Discovery?
I saw that a few weeks back.
Mucho COOL!!! B)
-
Don't you have that backwards shvarz? I thought I proposed the unmutable tags and you were the one who didn't like it.
Mmm, maybe. Yeah, my idea was called "UID" - for unique ID. I was proposing to insert non-mutable unique ID after each program in the DNA, then duplicate/delete these IDs together with insertions/deletions in DNA. That way we would have an easy way to align sequences that are evolutionary separated from each other.
-
Saving the DNA from within the program leaves the text file as a almost unreadable list of numbers and sysvars.
GET IT FIXED PLEAAAAAAAAASE!!!!!!
[EDIT]
I thought shen was talking about a serios bug in the save rotines again , I just realised he is probebly talking about that it writes everything in one calume. Not in the orginized way its displayed in the dna editor.
If you dont have time , dont bother (num py).
[/EDIT]
-
I'm trying to write a C++ program that'll analyze the syntax and attempt to render evolved DNA in human-readable form.
-
I think that rendering the dna in a more human format is only half the problem. The real issue is that, for better or worse, the dna is getting to be almost as complicated as real dna. The added complexity(above and beyond the complexity of most if not all other Alife programs) is always nice to see but at the same time it would be good to know what some of it really means.
I can almost envision a DB Genome Project someday :)
-
When I talked to some of my more technology-oriented friends, they told me that their degrees (like bachelor or master) are often granted for figuring out how a particular complicated instruments operates. What I see is some programming schools granting degrees for figuring out how some of the complicated bots operate :)
I already described somewhere the program that would make the job of figuring out what bots do much easier: a virtual "room" with a bot placed in it and toolbars showing values of all memory locations, stack and DNA. It would allow you to go through DNA one command or one gene at a time, highlighting memory locations read from/written to. If input is requested by a gene (say an "eye value") then user can input any value he likes. Something like that. I think PY liked that idea, but I don't know if anything is done along these lines.
-
Yes I like that idea and I certainly plan to add something like this.
Right now though, I'm waiting for Num to get the code for the next version sorted out.
He is making some MAJOR changes to the way everything runs. You won't see much of it as a user but from a programmer's perspective it is almost unrecognizable. I have a half made version of his code but I can't really do any work on it until he has it ironed out. There are bits missing all over the place still so it is hard to know where to start.
Once we get the format sorted then I can start on the peripheral stuff.
-
:drool:
(I knew it would be handy :) )