Part 4: Additional tools

This fourth and final part of Machine Learning in Action covers some additional tools that are commonly used in practice and can be applied to the material from the first three parts of the book. The tools cover dimensionality-reduction techniques, which you can use to preprocess the inputs before using any of the algorithms from the first three parts of this book. This part also covers map reduce, which is a technique for distributing jobs to thousands of machines.

Dimensionality reduction is the task of reducing the number of inputs you have; this can reduce noise and improve the performance of machine learning algorithms. chapter 13 is the first chapter on dimensionality reduction; we look at principal component analysis, an algorithm for realigning our data in the direction of the most variance. Chapter 14 is the second chapter on dimensionality reduction; we look at the singular value decomposition, which is a matrix factorization technique that you can use to approximate your original data and thereby reduce its dimensionality.

Chapter 15 is the final chapter in this book, and it discusses machine learning on big data. The term big data refers to datasets that are larger than the main memory of the machine you’re using. If you can’t fit the data in main memory, you’ll waste a lot of time moving data between memory and a disk. To avoid this, you can split a job into multiple segments, which can be performed in parallel on multiple machines. One popular method for doing this is map reduce, which breaks jobs into map tasks and reduce tasks. Some common tools for doing map reduce in Python are discussed in chapter 15, along with a discussion of how to break up machine learning algorithms to fit the map reduce paradigm.