The pole balancing code did not originate with me; I copied it from the prior-existing ESP source code for the experiment. I believe the ESP version may in...
Tom, did you every solve the issue with the time series and rtNEAT? Best Regards Ty ... which ... reproduce ... doing ... of ... best it ... being ... need a ...
Ty, did you try to somehow send me a message on compiling rtNEAT through this group? It seems somehow I got the message but it never got posted on the group....
Ken, I must have clicked on the wrong button. I have been having some problems trying to compile rtNEAT. However, today I stumbled across a solution. I run...
No I gave up and just built a static network but I would be interested if anyone has got it to evolve effective timeseries prediction. Tom ... -- PS Note this...
Hello all! I'm interested in using NEAT with Java, so I downloaded Neat4J as well as JNeat, but I could not make it really work for me. For JNeat I have done...
Cesar, would your friend be willing to give us permission for you to upload that version of rtNEAT to the files section of this group? I believe a number of...
Hi Max There is a bug in neat4j when you specify more than one output (Doh!). If you go to the sourceforge site for neat4j then there is a fix documented. ...
Please find included below a weblink to a short and useful reference regarding the importance of good practice in pseudo random number generation and...
Thanks for this. It's interesting to that George Marsaglia's KISS RNG is given the thumbs up, I used his XOR shift algorithm as the basis for my FastRandom...
In instances of a population being dumped onto a distributed processing enviroment for fitness evaluation, and a random input is required as part of the...
... Yes I accept that ensuring different RNG seeds across distributed nodes is critical. This is certainly a very good point, and one I hadn't considered...
Hello there. I am looking for any back-propagation algorithm that can work on networks with arbitrary topology such as these that NEAT evolves. All libraries I...
Hey Peter, I've never used it, but Backpropagation Through Time (BPTT) is a general method for applying backprop to recurrent neural networks. Here's a paper...
Hello- I believe Nolfi, Parisi and company did work a ways back combining evolution and learning in neural networks. I am not sure if those papers would help ...
Colin, this issue is outside my expertise but I agree that (I think) patterns in the sequences shouldn't really matter for the purposes of NEAT. There is no a...
... I totally agree. I just wanted to make the point that NEAT research performed to-date with potentially poor PRNG's most probably isn't lessened in value or...
A number of people have programmed backprop into NEAT. Chris Christenson did a Masters thesis on combining NEAT and backprop; a paper based on this work is...
Hi Ken, I am evolving time series predictors, in fact even a simplified version of time series predictors, where the network has to answer is the future value...
Dear all, I try to evolve a time series predictor where the input and output sensors have specific and meaningful positions in a Cartesian x,y,z space. As I...
Hi Christofer, The answer is yes, you can have any topology with HyperNEAT. You query the CPPN for every possible "from -> to" connection in the substrate. But...
Hi Peter, That means that I have to define a fixed network with hidden nodes from the beginning. The CPPN evolves, the phenotype is fixed. While only input and...
Christofer, only the nodes in the substrate need be fixed. You can allow HyperNEAT to determine how to connect them by imposing a threshold on the output of...
Peter, I believe that backprop can potentially improve the accuracy. It has been shown to work effectively with neurevolution in classification tasks in 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.
