Abstract
Neural Cellular Automata (NCA) models are trainable variations of traditionalCellular Automata (CA). Emergent motion in the patterns created by NCA has beensuccessfully applied to synthesize dynamic textures. However, the conditionsrequired for an NCA to display dynamic patterns remain unexplored. Here, weinvestigate the relationship between the NCA architecture and the emergentdynamics of the trained models. Specifically, we vary the number of channels inthe cell state and the number of hidden neurons in the MultiLayer Perceptron(MLP), and draw a relationship between the combination of these two variablesand the motion strength between successive frames. Our analysis reveals thatthe disparity and proportionality between these two variables have a strongcorrelation with the emergent dynamics in the NCA output. We thus propose adesign principle for creating dynamic NCA.