... thanks Ken. i will try Modular NEAT too however, and compare it to Hyper NEAT after that, i think on a big cluster working on small problem modular NEAT...
Hello Nikolai, I did some extensive testing with the modular NEAT code, it is strongly bugged. I tried to fix them but gave up after 6 Weeks. Also I found some...
Colin, I don't think there is any agreed standard set. However, I generally recommend not having more than one of each kind. For example, Gaussian is...
... Understood. Start with a minimal set and maybe introduce others at a later time if they show promise. E.g. I wonder about functions with abrupt changes in...
Hi all, I am involved in a modular robotics project (www.symbrion.eu), where we are developing robots that can hook onto each other to create 'organisms' that...
Colin, there really has never been a definitive study as far as I know on how different sets of activation functions affect performance. I am sure it would be...
Hi Ken, Thanks for your reply. I agree that step 2 is the more open. Intuitively, I prefer the interleaving approach as you do, so I'll try that first. As to...
All, I tend to agree with Ken Stanley. My only study on the effect of activation functions was conducted on a series of 100 double pole balancing experiments ...
Hi, i would like to check this idea with you guys. The problem has around 10,000 inputs, and only 3 outputs. So, in order to discover which inputs are more...
Hi Nikolai, I think that just lowering the add-node mutation rate should be enough, but why so many inputs? Is it a kind of image processing task? Maybe...
Hi All, I'm starting some HyperNEAT experiments trying various substrate designs. I'd like to be able to define them intuitively, e.g., in some XML format (a...
... it is possible to visualize networks with graphviz, together with gifview program from gifsicle tools for gif animation. here is an example of a network in...
Thanks Nicolai - that's great. Not a GUI, but still great :-). I know that boost::graph can read these files - have you (or anyone else) tried converting...
Evert, Yeah, I've re-architected NEAT in such a way as to be compatible for the Boost Graph Library - or any other graph library for that matter. I posted the...
Such an XML standard would be a good thing to have. While I agree that an expressive XML definition as you suggest is a nice thing, it does put the onus on the...
... Hey Evert, I wrote a 3d substrate visualizer for the HyperNEAT C++ implementation, but it isn't polished enough to release to the public yet. I've been...
Hi Jason, That sounds like a very useful tool - pretty much what I meant. What requirements does it have (i.e., ms windows or openGL)? Cheers, Evert...
Evert, I have been exploring creating models of systems and substrates for TWEANN neuroevolution in both UML and SysML, the latter being a profile of the ...
Hi Ken, Thanks for the clarification. As a follow-up, do you have any suggestions on how to handle mixed geometric and non-geometric data? For instance, if...
Hello, I'm having the hardest time trying to compile Hyperneat 2.6 on Fedora 11 using GCC 4.4. I already have zlib-devel packages installed, so I try and...
Hello, Browsing the archives, its obvious other's have worked on this problem using NEAT with varying degrees of success. I'm using the rtNeat C++ package...
Hi, Just a couple of thoughts. If you're training on a long period of price data then you should probably inflation adjust the prices, e.g. yahoo finance has...
Hi, i have trying to compile HyperNEAT under gentoo, and i have a linking error: master Hypercube_NEAT # make Linking CXX shared library ../out/libNEAT.so [...
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.
