The most important thing now is no longer debating whether differentiable programming should exist for Swift (because Swift + ML is sooo important!), but figuring out the best form that should land in the language!
—Richard Wei on Twitter
This short practical chapter is designed to introduce some deep learning–specific features of Swift for TensorFlow. Section 4.1 introduces the concept of tensor data structure which is essentially what neural networks make prediction from. After reading this chapter, you will be able to load the datasets (Section 4.2), write your own neural networks ...
