New error bounds for deep networks using sparse grids

Hadrien Montanelli, Qiang Du

We prove a theorem concerning the approximation of multivariate functions by deep ReLU networks. We present new error estimates for which the curse of the dimensionality is lessene…