ECJ-INTEREST-L Archives

February 2008

ECJ-INTEREST-L@LISTSERV.GMU.EDU

Options: Use Proportional Font
Show Text Part by Default
Condense Mail Headers

Message: [<< First] [< Prev] [Next >] [Last >>]
Topic: [<< First] [< Prev] [Next >] [Last >>]
Author: [<< First] [< Prev] [Next >] [Last >>]

Print Reply
Sender:
ECJ Evolutionary Computation Toolkit <[log in to unmask]>
Date:
Wed, 6 Feb 2008 14:51:27 -0500
MIME-version:
1.0 (Apple Message framework v752.3)
Reply-To:
ECJ Evolutionary Computation Toolkit <[log in to unmask]>
Content-type:
text/plain; charset=US-ASCII; delsp=yes; format=flowed
Subject:
From:
Sean Luke <[log in to unmask]>
Content-transfer-encoding:
7bit
Comments:
To: ECJ Evolutionary Computation Toolkit <[log in to unmask]>
Parts/Attachments:
text/plain (43 lines)
The CVS update was so large a few days ago that I think it's  
worthwhile mentioning what's in it for everyone.  I'm mentioning  
these because we could have introduced significant bugs and I wanted  
to warn people and ask for some testing.

If you can't get CVS and would like a snapshot before we post one (in  
a while), let me know and I'll wrap it up and mail it to you.

The big stuff:

	- We've made large modifications to ESSelection.  Previously your  
pipeline had to call it precisely once per generation of a child.   
Now you can have as many ESSelection objects in the pipeline, and  
they can be called multiple times -- they all select the same parent.

	- We've made big changes to various VectorSpecies and  
VectorIndividual subclasses, mostly getting rid of minGene and  
maxGene -- now there are only the arrays minGenes[] and maxGenes[].   
This radically simplifies the code at the cost of slowdown at  
initialization when it goes through and double checks that all the  
minGene/maxGene values are correct.  It also allows mixing different  
methods of setting gene min/max values (per-gene, per-segment, and  
globally).  This involved a lot of cut and paste and may have  
introduced bugs.

	- We radically revamped the distributed evaluation facility (the  
'eval' directory), to stomp a bunch of bugs.  This facility is now  
simpler, ultimately easier to understand, and more robust.  But I've  
not redone the documentation so it's probably a bit confusing for the  
moment because of the lack of docs.

	- We've modified the steadystate facility to work better with the  
distributed evaluation facility.

	- I changed Output.systemMessage() to be on the low end of the  
verbosity totem pole to help the "ga within a ga" discussion that's  
gone on recently.

Once again, we'd be very grateful for testing, particularly in the  
Vector package, where I have the most fear of introduced bugs.

Sean

ATOM RSS1 RSS2