Hi all, Having read the recent CPPN papers from Ken's group it seems to me that this is potentially a large leap forward for neuro-evolution. So for me an...
Please find included below (4) useful libraries for NEAT C++ code development: "CString-clone Using Standard C++", O'Leary, Joe. A Drop-In replacement for...
Hi Colin, yes I agree that visual domains would be interesting to revisit. More generally, domains with potentially many inputs and/or outputs that are...
Hi Ken, The good people over at http://galaxyzoo.org/ announced today that each galaxy in their database has been categorized by 30 people (on average). Sounds...
Hi all, I remember that we discussed one issue about visual recognition with HyperNEAT (CPPNs) and one particular problem arised and this the problem of...
Thank you Ken, it is really fun thing to play with :) What about using LIs (leaky integrators) to produce smoother movements and stuff? I think this could make...
Hi, all. My idea for using NEAT/rtNEAT is for my honours project at uni where I, hopefully, will be using NEAT/rtNEAT in a game. One major idea I have had...
Hi Tom, This is a great idea for an application of NEAT, but this thread is about CPPNs/HyperNEAT, a new kind of indirect encoding, not the original NEAT which...
Hi Peter, Thanks for the ideas. I can see now that I really need to go back into the message archive and acth up on some of the past CPPn discussions. e.g. I...
For those interested in CPPNs and HyperNEAT I would also suggest starting out with a domain that is not too complicated just to get a little practice since...
(Yahoo seems to have lost my first response to this post so in case it appears later I apologize for responding twice..I sent it out yesterday and it still...
I agree, but I need to mention that some domains do not require setting up a substrate at all. Consider the drawing of a picture and transforming it to a...
Peter, you are right about the interpretation of both -1 and 1 as white. This tradition goes back to Mattias Fagerlund's DelphiNEAT- based Genetic Art. One...
Stanley, I'd like to know if you've analyzed the Cellular Encoding approach while developing NEAT and what's your view about Cellular Encoding X Direct ...
Rafael, our 2002 paper on NEAT does include a comparison with Cellular Encoding (CE): http://nn.cs.utexas.edu/keyword?stanley:ec02 Our 2003 "Taxonomy for...
I should add that as far is direct vs. indirect, I would go with indirect on very complex problems, but CE is probably not the best indirect encoding today. ...
Hi all, I've been looking at the jneat code, specifically Genome.mutate_add_node, to try to understand how innovations work in this case. If the connection to...
Hi Peter, Did you eventually modify SharpNeat to support validation data sets? Or does anybody know if there any Neat implementations/code snippets etc out...
Hi Keith, I wrote the CodePlex version of SharpNEAT one week last summer. It's essentially the same as the official SharpNEAT (Sourceforge version), but it has...
Hi Rob, The solution to this problem is to look in the genome for a node ID that matches the newly created one. If such a node is present in the genome, then...
Yet another "I can't get XOR to work" post :( I'm trying to implement a NEAT package in ECJ. To test it, I'm running XOR with fixed topology, but my runs only...
Actually, depending on your perspective, both of you are correct. If a hidden node is limited to one input and one output it takes two hidden nodes for XOR....
OK, so my hidden node has three input connections -- from the two inputs and the bias node. The output node has four input connections, from the two inputs,...
Hi Rob, I used SharpNeat to search the fixed topology you described (so weight mutation only) with the same activation function and I came up with a working...
Colin, are you sure that SharpNEAT flushes the network between the 4 input examples of XOR? I think I remember someone telling me that there was a some kind...
... You may be thinking of ANJI. That happened with us when we first implemented XOR. Oh, look, it solved XOR...yay! Oh wait, the error is huge when we ...
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.
