deep learning methods continue to have a decided impact on machine learning,
both in theory and in practice. Statistical theoretical developments have been
mostly concerned with approximability or rates of estimation when recovering
infinite dimensional objects (curves or densities). D