=== INSTRUCTIONS === https://sulfurous.aau.at/#200358027 (Use turbo mode for fast learning.) If the population doesn't seem to be improving, try restarting the project. If the bestFitness gets to 3000+, then it's on the right track. === UPDATE === You can now train the best species in sulfurous! Once the bestFitness (cloud variable) gets higher in sulfurous, press space and comment what the code is. Sulfurous can't update the cloud variable, but I can do it manually. Press 'C' to set the weights of the first two species to the best set. === MORE INFO === Inspired by: http://www.ai-junkie.com/ann/evolved/nnt1.html GA + ANN stands for 'Genetic Algorithm + Artificial Neural Network', as in a way to train a simulated population through a genetic algorithm (think evolution). My favorite part of this is that the neural network structure is easy to adapt. Simply edit the (NN) list, with each item being a layer and the number of nodes in that layer. Be careful editing the first and last layer though, as those are the inputs and outputs (you'll have to edit some custom blocks if you want to do that). If you want to adapt the environment itself, you'll have to edit the custom blocks on the left side of the editor and change any custom blocks on the right if you are editing the (Population) list.
