conda create -n evergreen python=3 conda activate evergreen conda install numpy tensorflow-gpu=1.15 git h5py pip install gym pip install pygame git clone https://github.com/adamjcalhoun/Evergreen.git
Train on two different patch types - coarse, dense, and all pairwise combinations of small/large Train several models - dense, LSTM Also: odor memory? Show chemotaxis ability Show foraging Show representations
Dense network -> 1 or 3 hidden layers -> reward hunger or reward odor -> coarse (patchy) and dense patches (dense patches should have equivalent amounts of food)