Neural Networks: Depth Trumps Width in Analytic Function Approximation

New research shows depth is more important than width in neural networks for approximating analytic functions. This shift in focus challenges traditional thinking.