From e6ab4a8ed100d5d5b7611c74cf3ccd556f1f1d71 Mon Sep 17 00:00:00 2001 From: Paul Oliver Date: Thu, 29 Feb 2024 19:04:34 +0100 Subject: Initial commit --- README.md | 22 ++++++++++++++++++++++ 1 file changed, 22 insertions(+) create mode 100644 README.md (limited to 'README.md') diff --git a/README.md b/README.md new file mode 100644 index 0000000..8e68cec --- /dev/null +++ b/README.md @@ -0,0 +1,22 @@ +rtES-HyperNeat (Real Time Evolving Substrate Hypercube based Neuro-Evolution of +Augmenting Topologies) is a real time neuro-evolution algorithm based on the +papers written by Kenneth Stanley and other researchers at the EPLEX group +([http://eplex.cs.ucf.edu/](http://eplex.cs.ucf.edu/)). More specifically, +rtES-HyperNeat tries to blend the existing ES-HyperNeat +([http://eplex.cs.ucf.edu/ESHyperNEAT/](http://eplex.cs.ucf.edu/ESHyperNEAT/)) +algorithm with rtNeat +([http://nn.cs.utexas.edu/?rtNEAT](http://nn.cs.utexas.edu/?rtNEAT)), a previous +version which allowed real time evolution of agents on a virtual field. + +The main problem with trying to make ES-HyperNeat real-timed was that the CPPNs +(genotypes) had to be converted into functional Neural Networks (phenotypes) on +a single time-step. This was proven to be impossible, so, as a solution, the +library handles these transformations on separate execution threads. This way, +the algorithm flows seamlessly. + +Although several implementations existed for HyperNeat, I've decided to code +mine from scratch, based only on the original papers by EPLEX. My goal is to +construct a simple, easy to use, well documented and tutorialized C++ library +that will allow any user to evolve Neural Networks for any imaginable task. I +also plan to apply this library to my Guppies experiment once its finished (see +my channel for more info on Guppies). -- cgit v1.2.1