On the minimax optimality and superiority of deep neural network learning over sparse parameter spaces

Open in new window