We explore how a baseline fully connected neural network, a long short-term memory neural network, reservoir computing, and a Lagrangian neural network compare on the task of forecasting chaotic motion, namely the prediction of the path of a double pendulum.
Leave a Reply