Hello all- There are frequently conversations on this list about what parameter settings to use for evolutionary runs. At the CEC conference this year I saw a...
The paper is also available online: http://www.cs.vu.nl/%7Egusz/papers/2009-CEC-tuning-methods.pdf Not unsurprisingly, the CMA-ES comes out on top again... ...
I am pleased to announce that GAR is now released and available at: http://gar.eecs.ucf.edu As some of you know, GAR can be played single player or multiplayer...
Hi, I'm working with neat4j but I need to make it stoppable in order to resume an evolution later. I did it saving the serialization of the population object....
or, having re-read your post properly it is probably an issue with a mismatch between the serialised population and the innovation database which tracks the...
I downloaded it and installed it, but couldn't run it. It gave me the "Send/Don't Send" error message and that was it. I have all other components installed,...
Peter, that's strange I have not heard of anyone not being able to run it who had everything else correct. It sounds like you have the right software and...
Hi Neat people, Some of you might be interested in a deadline extension in the forthcoming IROS workshop on Evolutionary Design of Robots taking place next...
Hi, as I understood, the innovation number that is assigned to each genome is valid only to run neat on one computer. For example, if i run neat on the same...
Is, i was already aware of that bug. What i did is saving the serialization of the innovationDatabase along with the population. It's seems working......
Hi, Yes, typically a NEAT implementation will have some means of comparing new connections (and sometime's neurons) with all of the connections currently...
Hi everyone, Did anyone else find it problematic that the NeuronGeneId was generated just like the InnovationId for the ConnectionGene? In my implementation, I...
Hi, ... Not especially but I did decide to use two separate incrementing counters, one for neurons and one for connections. The two types of gene have their...
Colin, Actually, there is more to the story. Indeed, I am using a highly distributed form of TWEANN based on NEAT run on hybrid, heterogeneous computer...
Nikolai, there is actually an interesting alternative solution to this problem: Instead of using a counter for the innovation number, you can use a very...
Hi Ken, ... It did occur to me that the connection IDs are redundant information in representing a graph. Say we are using 32bit integers for the node IDs,...
... have you guys considered for NEAT v2 a gene to be not just a link, but a group of links ? And even more, a gene to be a well defined large network ? I have...
Hi Peter Great demo. Can I take you up on your offer to release the source code for the demos? I'm interested in co-evolutionary approaches with an agent and...
Hi everyone, I am working on a project that needs a substrate of more than 2 dimensions. Looking at David and Ken's GECCO'08 paper, I would have expected a 3d...
Nikolai, the idea of having a gene encapsulate a number of components is similar to the idea in Modular NEAT from Joeseph Reisinger a few years ago: ...
Wesley, the main insight motivating r(x) is that it aligns the agent geometry with the team geometry. That is, the left side of the agent is on the same side...
Hi all, Just wondering what are currently considered to be the standard set of CPPN activation functions. I figure they are: Sigmoid Gaussian Abs Sine Although...
Also it just occured to me that abs, linear and one's complement might cause problems because they have an unbounded output range - thus signals propagating...
... 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...
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.
