Neural ordinary differential equation control of dynamics on graphs

We study the ability of neural networks to calculate feedback control signals that steer trajectories continuous-time nonlinear dynamical systems on graphs, which we represent with ordinary differential equations (neural ODEs). To do so, present a ODE (NODEC) framework and find it can learn drive graph toward desired target states. While use loss functions not constrain energy, our results show, in accordance related work NODEC produces low energy signals. Finally, evaluate performance versatility against well-known controllers deep reinforcement learning. generate controls for more than one thousand coupled, ODEs epidemic processes coupled oscillators.