Modulating chaos in spatiotemporal systems based on deep reinforcement learning

A reward function based on the moving largest Lyapunov exponents relying on a fixed-length set consisting of recent states evolving over time, which only involves a small subset of the system’s variables is designed.