That’s it for this introduction to machine learning. Congratulations for making it this far! You should now understand what machine learning is, know how to use it, and have a sense about how it works. Don’t worry if some parts of this book are still fuzzy to you. There are a lot of machine learning concepts and it takes time to fully understand them. Also, you don’t need to understand everything to start applying machine learning in a useful way (in fact, you might discover that almost all practitioners have knowledge gaps).

Okay, so what should you do next? The most important thing is to apply what you learned in one way or another. The exercises in this book help, but nothing replaces real-world practice. If you just want to understand what machine learning is, this might mean discussing the machine learning projects around you. If your goal is to practice machine learning though, you should start working on some machine learning projects. The best scenario would be to find a project that is actually useful for you or your organization. Alternatively, there are plenty of online machine learning competitions that you can participate in.

While this book covers the modeling and testing aspects of machine learning, there are other things to consider when working on a machine learning project. Here are a few generic pieces of advice. The first thing to do is to make sure that the problem is well defined and actually useful. If that is so, the next step is to figure out if the problem could be tackled using a simple classic program instead. Classic programs have several advantages over machine learning solutions, such as easier debugging and maintenance, so they should not be overlooked. If traditional programming is not practical, you should then figure out if data is available and if a machine learning solution has a chance at succeeding. If that is the case, you can actually start the machine learning project. You would typically start by exploring the data, training a simple baseline, and then doing a bunch of experiments (trying various methods, trying various feature extractions, and obtaining more data). Each experiment leads to the next experiment, so it is best to iterate quickly (it typically helps to use smaller datasets). Finally (this was already mentioned in the book but it is important), it is essential to use good validation/test sets. The validation/test sets should be disjoint from the training set and be as close as possible to the in-production data. This is a difficult but necessary part to ensuring that your model is actually working.

Besides practicing, you might want to explore some of the areas of machine learning that are not covered by this book. One of these areas is reinforcement learning. Reinforcement learning is currently mostly a research field, which is why we did not cover it, but its uses are growing, and it is a fascinating topic in itself. Another area that was not explained is how the unsupervised learning methods work, which you might be interested in. Finally, an important area that was only briefly mentioned in Chapter 5 is called responsible AI, which concerns the set of methods and techniques used to make sure that machine learning models are fair and properly used. Responsible AI is needed whenever a machine learning model is applied to humans, such as for medical or hiring decisions. If you ever work on such applications, it is important that you learn about this. Aside from current areas, there will also be new methods and techniques that emerge because machine learning is a rapidly evolving field (and this book was written in 2020/2021). Fortunately, you should now possess the necessary vocabulary and overall understanding to enable such continuous learning.

That is it. I hope that you enjoyed reading this book, that you learned interesting and useful things, and that you will take part in this ongoing machine learning revolution.