Chapter 5. Math
If the extent of your ML work will only ever consist of running keras.fit or cloning existing implementations, you probably don’t need math. There are many courses and books that promise you machine learning mastery with little or no math at all. If that’s what you’re looking for, feel free to skip this chapter.
However, some mathematical background will be helpful to the following.
- evaluate the tradeoffs of different algorithms and choose the ones that work best for your problem
- debug your models if something goes wrong during training
- make changes to improve your models, either in performance or efficiency, even if nothing goes wrong
- explain certain aspects of your model performance
- create new models.
This section covers the following branches of math that are important in ML: algebra, probability and statistics, dimensionality reduction, and very little calculus and convex optimization. This list is far from exhaustive. For example, graph theory, logic, topology, and other mathematical branches occur frequently in ML but aren’t included here.
If some characters seem to be missing, it's because MathJax is not loaded correctly. Refreshing the page should fix it.