Hello all, The version of novelty search used in the novelty search paper is now available at http://eplex.cs.ucf.edu/software.html. Post here or send a...
Julian, Thanks for the further insight on CMA-ES. I also think there may be a way to combine it "properly" with a variable-dimension encoding. Somehow the...
Ken, I presented our first memetic climber paper at WCCI yesterday, and one of the people in the audience was Nils T. Siebel, coinventor of the EANT-II...
Julian, I am sure you are enjoying WCCI and presenting your work there. I'm looking forward to GECCO next month. Are you planning to be at GECCO? I have...
Hello, This email message is a notification to let you know that a file has been uploaded to the Files area of the neat group. File : /novelty_nevh.rar ...
neat@yahoogroups.com
Jun 9, 2008 12:24 am
4133
Wow, this looks really cool Petar. I thought it would be awhile before we saw HyperNEAT and novelty search combined. I ran it while I was working on other...
Hey I am glad that you like that stuff, it was around in my computer a whole week but I decided to upload a binary just before a day. Let me explain the...
Nice work, Peter, it's a very nice demonstration. I also saw the humanoid shuffle backwards. It does seem easier to do that than go forwards. My guess is...
Thanks, Ken. Novelty search really does much better than fitness search in this domain. I agree that HyperNEAT allows too much recurrence in this particular...
Thanks a lot to Petar C for his novelty-based nevh, its really coming along nicely. In the past while I have been thinking of how to best exploit parallel ...
Although the papers on hyperneat propose some unique ways of representing the substrate geometrically, I got the impression that the resulting generated net...
Well the substrate in NEVH is not a simple single layer perceptron thing, but a Continious Time Recurrent Neural Network (CTRNN). Such networks with enough...
Interesting points Petar. It was my understanding from the original hyperneat papers that points on the substrate correspond to nodes in a potential network....
Petar, I understand what you are saying, and indeed you can do it exactly as you described using rtneat. I should have clarified, my implementation is not a...
Peter B, just a few notes on substrate layouts: If you do designate some nodes as inputs, some as hidden and some as outputs, then given those designations it...
Hey Peter, I've been working on an MPI implementation of novelty search, so it is interesting to hear other people's ideas about parallelizing the algorithm. ...
I was reading about this http://www.umiacs.umd.edu/~vikas/Current_research/scalable_machine_learn ing/scalable_machine_learning.html Apparently what I glossed...
Hey all, I've just recently started working with leaky integrate-and-fire (LIF) neurons, and since I don't have any prior experience with them and I'm relying...
Derek, A friend of mine has implemented an InF module (as well as the Izhikevich spiking model) for our NEAT Python project. I'm not very familiar with them,...
Hey Cesar, I guess your friend isn't on this list? Has she actually run any simple simulations using this implementation? My advisor is a big fan of...
Hi Derek, ... She was on the list, I don't think she is anymore (need to talk to her!). Anyways, the Izhikevich model has a simple working example. If you have...
Cesar, Could you tell me a little more about your NEAT Python project? I have heard a bit about it but I don't know what it's being used for or whether it is...
Hello Matt- Apologies for the delay in responding to this. I flagged it as something to be thought about and replied to, and then it scrolled off my screen...
Hello all- This paper featuring HyperNEAT will be presented at PPSN 2008 (Parallel Problem Solving from Nature) in Germany. Some of you may find it ...
Hi Ken, ... We started our own Python implementation because we felt that's easier to extend and develop. I've also taken a look at PyNEAT (another older NEAT...
Very cool stuff Cesar. How does it compare speed-wise to other releases? Cheers, Jeff Clune Digital Evolution Lab, Michigan State University jclune@......
Thanks Jeff! Well, it's faster than I thought it would be considering it's a Python application. The neural network module has some C++ bindings to python, 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.
