Hello group, I'm very interested in experimenting with NEAT but I'm having some problems getting it to compile. I'm using GCC 3.4.4 and I tried applying the...
Greetings all, Just to let you know that the reason I've been a bit quiet lately is I was busy finding myself a new job. That's all done now, and I started...
Out of curiosity, is there any variability in the fitness evaluation? Also, when you graph the fitness per cycle, graph the complexity, too. If the complexity...
... Well good luck with that, looks 'your cup of tea' anyway. I looked for work this time last year but was thoroughly disheartened by the type of work...
The problem looks like it has to do with compiling gtk and Guile. As time goes on, the way NEAT compiles and references these libraries is probably becoming...
Hi, ... That's probably a very cool idea, we could build in terms for the various things we already identified as coming in 47 different flavours, e.g. I had 8...
... Sure, but I just think that some of the terms in use right now are a bit vague and confusing. ES is a good example and in general I have to go and look...
... All this is correct, but not the cause of my slowdown. I meant to explain this in the last mail, in fact it was the whole point of what I was saying, but...
... We handled this in our floating eye shape evaluator by giving each individual the same number of CPU cycles (appoximately). So, the larger the neural...
... I suppose though that evolution in general doesn't normally find perfect solutions, just solutions that are good enough. When you have a very organic...
Hi, I've been thinking about speciation a bit more, and in particular in the context of my current SLEAT situation... The problem I have, is that (this is with...
Hi Ian, Interesting post. I Just have a couple of points/thoughts... ... Hmm maybe, but. If it is fitter it will take up a larger proportion of the population,...
... Sorry, I wasn't clear what I meant. When I called this "alternative speciation" I meant as a strategy to use entirely instead of the current approach....
... Understood. ... One aspect of this sort of approach may be to more thoroughly examine the fitness space around a given species. The current mechanism of ...
... Was just thinking about this point and about dominant and recessive genes. I'm wondering if maybe we could incorporate this type of idea into our...
Hi, In Nature (tm) recessive and dominant genes come about not by turning one of other of them off, but because one of them does something and the other does...
Hi Ken, I was finally able to get it to compile with GCC 3.4.4. The primary problem (after I applied the patches) was that I was trying to compile neat, not...
Hi Tom, I'm glad to hear you could finally compile it. Is it possible you could share the Makefile and any changes you made to the source to get it to...
Hello, I'm currently writing an implementation of NEAT in Ruby. When I'm finished, I plan to evolve neural nets to control a CYE robot, taking input from a...
... I'm afraid that neuroevolution requires a computer simulation of the system to be controlled by the ANN. If there is any other way of doing it I would be...
Well, it doesn't require simulation, but simulation lets you do a whole lot more runs in a whole lot less time. At a conference presentation at GECCO, Dario...
I had not considered that there might be situations where hardware fitness testing is practical. For the usual robots that we are familiar with it is usually ...
Interestingly, since Thomas's goal is to test the roving eye idea on a real robot, the size of the robot doesn't really matter. The issue in this case is the...
Hi, So, further to my last post about alternative speciestion, I wanted to discuss the case where you have two species and want to merge them back together, in...
Hi, Does anybody have any thoughts on the causes of bloat...? e.g. the reason I ask is because, however much I push the probability of delete above that of...
Ian, What makes it bloat, and not useful complexification, is a disproportionate increase in size without a corresponding increase in fitness. So I assume ...
... As well as additive mutations, crossover also results in structure being copied throughout a population and quite possibly faster than deletion mutations...
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.
