Traveling waves of neural activity can be used to efficiently store and update the state history required for learning history-dependent dynamical systems in recurrent neural networks.
