Julian, thanks for sharing your recent work. I hadn't been aware of that paper and it is indeed an interesting study. I'm guessing that it probably has not...
Well actually the NEAT parameters may be variable. For example stagnating species can increase their weight mutation rate (as a last chance to discover a good...
Thanks for your comments, JT and Ken! JT: Yes, NEAT contains a similar idea, as part of a rather complex algorithm. We're trying to boil it down to the bare...
I'd like to encourage you all to consider participating in the WCCI car racing competition, which is about learning controllers for racing cars. In last year's...
Julian, here are a couple other questions/comments: I was curious why you chose to call it "memetic?" Usually I think of memetic as meaning taking the result...
Isn't this an awful lot like the cascade-correlation algorithm, in which hidden nodes are added incrementally and the network is retrained after the addition...
Ken, It might be that word memetic is indeed misleading - we just didn't come up with a better name for it! "Multiple timescale evolution" doesn't really sound...
Derek, Yes and no. Probably more no than yes, when I think about. Cascade correlation only grows nets, whereas the memetic climber cuts connections as well,...
... form of ... combining topology ... weights. ... obvious next ... I think that some people did see themselves as taking the next step when neuroevoltuion...
... Actually, the big difference is that we're doing it at different timescales. This is really the whole reason it works. The simplest extension is to mutate...
In instances in which fitness evaluation is non-trivial, effective utilization of computational resources can translate into broader successful applicatility...
I see what you're saying- no one thought to really explore the interaction of the two different time scales. But don't most of the older TWEANN neuroevolution...
... Apparently... I'd really like to know if the stuff is out there somewhere. It would be embarrassing to go and publish more papers on algorithms based on...
Hi everyone. I've been looking at the food gathering experiment discussed in the paper "A Novel Generative Encoding for Exploiting Neural Network Sensor and...
Hi, I wrote that paper with Ken and the code your looking for can be found here: http://eplex.cs.ucf.edu/index.php?option=com_content&task=view&id=17&Itemid=32...
Thanks for the quick response! I did not realize it was in the C# version. ... http://eplex.cs.ucf.edu/index.php?option=com_content&task=view&id=17&Itemid=32...
Thanks Julian, that explanation made sense. I see the distinction now. ken ... on ... have a ... rate? In ... time ... implicitly? ... precisely ... types of ...
Jason Gauci and I are pleased to share with you new results with HyperNEAT in our new paper to appear in the Twenty-Third AAAI Conference on Artificial...
... Cool, I enjoyed the paper, thanks Jason and Ken. I have a few questions: * Was the 4in/2out CPPN arbitrarily chosen? * Why not a 6in/1out (xyz, xyz, out) -...
Good paper! It is now clear that HyperNEAT is indeed a very powerful type of NE algorithm. I believe that CPPNs are the best abstraction of development found...
Hey Stephen, Great questions! We'll have to think about some of them for awhile but here's a few answers: The reason why I chose to use a separate output for...
Jason, Couldn't you write some screen scraping code to automate the information extract and play for a checkers website? I've done something similar for...
Another interesting thing about having a CPPN with two outputs (one for each layer) is that theoretically you can get two weights out for each single complete...
Regarding the number of connections, there are two things to note: First, HyperNEAT uses all 64 input nodes, even the ones which cannot legally contain a...
Do you allow direct input-output connections in this case? ... method, ... as ... are ... like ... performance ... the ... to me? ... process? ... You...
These certainly are impressive results in a very cool domain. I've know about them for a while, but just recently got to read exactly how the CPPNs work in...
Sorry for this double-post, but if you don't allow direct input- ... from one hidden node to another one in the same layer), then you have made assumptions for...
Yes true, that is something to consider. Direct connections are often useful as well and without z there would need to be perhaps another output for those....
Hi.... iwant to use matlab_neat for the classification of Breast CancerData.... can you please help me modifying the code so that i may use itfor the...
Ali, it's unlikely that someone would volunteer to modify code for you, but perhaps if you gave more information on your task someone could offer advice. ken ...
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.
