Time discretization invariance in Machine Learning, applications to reinforcement learning and recurrent neural networks
While computers are well equipped to deal with discrete flows of data, the real world often provides intrinsically continuous time data sequences, e.g. visual, sensory streams, time series, or state variables in continuous control environments.…