Hi, I have extended the class GP::Individual to included genealogy information, and as part of this I have altered the way individuals are stored (they now...
I find that when the max fitness in the population degrade, and I use hallOfFame size >1, the degraded worse max fitness individual in the current population...
Jianjun Hu
hujianju@...
Jun 5, 2003 4:10 am
199
Hi, I know the problem you mention. This is one of the flaws found in PipedInputStream. I will replace completly the XML parsers for next version. This will be...
... The algorithm, as I intent to implement it would not behave as you mentioned. The hall-of-fame will always keep the N different best individuals of the...
Hi, christian: Exactly!! There is no problem when the size of halloffame is 1! and problem comes when size is >1. I will test your new code soon. I think for...
Jianjun Hu
hujianju@...
Jun 5, 2003 6:59 pm
202
Hi Jianjun! I made a review of the Deme::updateHallOfFame method. I confess, I found 2 bugs in the code that may explains the problems you encounter. These ...
Hi, I have followed up the last mails about updating HallOfFame and I would like to understand some things: I have looked into the best of the best in the...
... The HOF is sorted in the descending order. So the best individual is the one at the index 0. The HOF uses the method Object::isLess and Object::isEqual for...
Hi, buddies: I am wondering currently how beagle support strong type GP. It's a pity that all current GP examples are not strongly typed. In crossover, I...
Jianjun Hu
hujianju@...
Jun 12, 2003 8:05 pm
206
oops. I found that the strongly type is implemented in an afterwit way: Blind crossover + validation. A much more efficient and simpler way is just do the...
Jianjun Hu
hujianju@...
Jun 12, 2003 8:20 pm
207
Dear Christian: As I am trying to refer Beagle experiments for my paper, I found that some implementation details are slightly different from other packages...
Jianjun Hu
hujianju@...
Jun 17, 2003 7:14 pm
208
Hi Matthew, Sorry for answering such late, I was on vacation. ... You are right. I don't remember a reason why I didn't do the check, so I suppose that forgot...
Hi Jianjun, ... Yes, it is done so. ... It was a deliberate choice I did when I developed the GP framework. I had two ways to implement constraints / strong...
Hi Jianjun: ... I agree that in the generational evolver, the operators are applied on the population in a quite different way than most other packages: the ...
hello all. i am using symbreg from the gp examples to model econometric data such as GDP (as koza described in (i think) volume I). the two questions i have...
... I guess you are talking about the member variable mY. And you're fairly much right, it is a vector of Beagle::Double. But vectors and arrays differ a...
Hi, ... You can use the following expression, which would work as you wanted. y[i].getWrappedValue() = MyDoubleArray[i]; ... The best of run GP programs are...
Hi again, It is interesting to see the perception are different from one person to another. I don't mean that I catched the question right, answers of Matthew...
... I've never thought about that. You are (of course) right---it's quite an interesting use of getWrappedValue(). You could also write y[i].setWrappedValue(...
... Yes, this is right. I'm quite used to getWrappedValue(), but in this context, it is more appropriate, cleaner, and OO to use setWrappedValue(). Anyway,...
Hi all, I'm not sure about how FitnessKoza works. As far as I concerned, I understood that the the 4 parameters used as input are the parameters that will be...
Hi Dan, I could hardly gave you any comments on your code, as I don't understand precisely what you want to do, from a GP, high-level point of view. Are you...
Ah! I already has executed the code with no problems, but the Hall Of Fame is showing best individual that doen't refelcts the truth. Is there any problem to...
... With the FitnessKoza, the individuals are selected only using the normalized fitness measure. The other ones (standardized, adjusted, raw, hits) are not...
Hi Christian, I already saw the tableau. The Goal of the program is to find a function that reflects aproximately a "Threshold" for taking decisions in the ...
Hi again, Ok, as I could understand, you are doing a kind of symbolic regression, where you test your individuals on several data sets. So the normalized ...
I am wondering is there an hallOfFame for the Vivarium when we have multiple subpopulation? If so, how can we access it? If not, this feature is strongly...
Jianjun Hu
hujianju@...
Jun 23, 2003 4:12 pm
225
... No, there is no hall-of-fame in the vivarium. Yes, this is an interesting idea. I'll see what I can do for next release. christian ... -- ...
Having problems with message search?
Fill out this form to ensure your group is one of the first to be migrated to the new message search system.
