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 : /NEVH_1_1.zip ...
neat@yahoogroups.com
Jul 2, 2007 11:39 am
3441
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 :...
neat@yahoogroups.com
Jul 2, 2007 5:25 pm
3442
I am having trouble in two main areas: fitness evaluation and Q- learning episodes in the NEAT+Q algorithm. After two generations the "max_fitness" drops to...
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 :...
neat@yahoogroups.com
Jul 2, 2007 11:44 pm
3444
Ken/Cesar, I have been working on my own version of NEAT in Python, but it is still in a fairly primitive state (many bugs and some of the more recent features...
I just read about Kenneth Stanley's NERO and rtNEAT, which is both infuriating and exciting at the same time. I was trying to establish an efficient method of...
Ken I sent Cesar the first version of my python neat stuff. If you still have any of the later versions with checkpoint/restart, etc feel free to release it ...
Unfortunately I don't have experience with the NEAT+Q implementation so Shimon is probably the best source for these questions. Hopefully Shimon will answer on...
Thanks for the welcome ;) There seems to be a number of possible directions to further NEAT, things that I can't find mention of. For example - has temporal...
I modified the NEAT Matlab code to perform the NEAT+Q algorithm, it doesn't quite work yet though. I spoke with Shimon and he basically told me that the...
Temporal delay and signal decay are generally present in leaky integrator neurons, a.k.a. Continuous Time Recurrent Neural Networks (CTRNNs). Tyler Street...
That sounds like good advice. Fitness in NEAT indeed must be scaled to above zero values. Noisy domains present a problem for any method and require many ...
Thanks for that, I'll have a look at their research to compare my ideas against. Your response, in itself, was quite educational (regarding the classifications...
Hi Brian, thanks for sharing your python NEAT implementation. Currently I'm modifying Simmerson's Java implementation (NEAT4J) to fit my needs but my interest...
Hi all, I'm wanted to let everyone know that I've finished an initial release of a parallel (distributed really) implementation of SharpNEAT and it's available...
Thanks Wesley. To clarify we conversed briefly off list about bringing the new functionality into the SharpNEAT release on sourceforge, but as I prefer to ...
Hi there, I've been experimenting with two different methods for neural network update in NEAT4J. I've used the same settings as Matt's: tahn(x) activation for...
Hi, I am trying to compare a Reinforcement Learning algorithm against NEAT, and I'm trying to test NEAT in a simple domain. Inputs = 1. Outputs = 4. ...
Hi Cesar Your results are very interesting. Is it possible you could send me your synchronous update code, and I will integrate it into NEAT4J, also, does it...
Cesar, I'm not sure if I understand completely how your activation method works, but it sounds like you may not be flushing the network between the four trials...
Hi Mridul, it looks like what you are trying to do is just cause NEAT to assign a weight of zero to every weight. Is that the goal? What inputs are you...
Here is a cool NEAT application that I thought people might find interesting. It lets the user interactively evolve dancers. The link leads to videos on the...
Hi, I think I found the bug. I converted Matt Buckland's code to Java because of project requirements, and I've probably done something off in the conversion....
... It works multiplying a connectivity matrix by the state vector of the network (synchronous updating like in Hopfield networks). ... That's exactly what's...
Hi Matt, The activation method is still in development and not well documented. And I'm not sure how well this method works for supervised learning tasks like...
Hi Ken, This application looks really cool :) It is like the interactive particle effects, drum beats and chords, but this time dancers :) It is interesting,...
Thanks Peter, we are looking forward to releasing it. Yes, they are basically ragdolls that are being controlled by neural networks evolved with NEAT. The...
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.
