Dimension-independent bounds on the degree of approximation by neural networks | IBM Journals & Magazine | IEEE Xplore