Exploring the Unreasonable Effectiveness of Mathematics in Deep Learning

The chapter delves into the unreasonable effectiveness of mathematics in explaining physics and draws parallels to advancements in deep learning with the availability of more data and powerful computers. It discusses the computational complexity of algorithms in computer science, contrasting combinatoric complexity with the order n complexity of learning algorithms for deep networks. The chapter also highlights the development of massively parallel deep learning chips and the shift towards hardware that can perform tasks in parallel, akin to how the brain functions.