In this chapter, we discussed a number of practical matters related to ML applications in production. Learning ML algorithms is, of course, central to building an ML application, but there's much more to building an application than simply implementing an algorithm. Applications ultimately need to interact with users across a variety of devices, so it is not enough to consider only what your application does — you must also plan for how and where it will be used.
We began the chapter with a discussion about serializable and portable models, and you learned about the different architectural approaches to the training and evaluation of models. We discussed the fully server-side approach (common with SaaS products), the fully client-side approach (useful for sensitive data), and a hybrid approach by which a model is trained on the server but evaluated on the client...