I told you on my last reading list article that I happened to bring “Thoughful machine learning with Python” from the library, just for fun, a look outside the bubble…
When I started to read that book, I realized that this was not an introduction to machine learning, but something like the “Clean Code” of the machine learning world, discussing style and practices for people who already know what all of that is about.
Since I have no background in machine learning at all, I put that book to the side for later and went hunting in the library once more, looking for the actual basics.
As the result, I’m reading “machine learning for dummies” this week. Having reached page 150 now, I want to give you a short interim assessment:
I don’t like it too much as a beginners book. As I said, I reached page 150 – and I haven’t seen any example or even a single line of code related to machine learning in that book. Compared to the “Head First” series, that makes you code running examples on the respective topic from page 1, that “for dummies” is disappointing. And I remember having had the same impression on a different “for dummies” 10 years ago.
Putting that aside, here are three facts I learned so far, which were both new to me and very valuable to raise my awareness on the topic of machine learning:
- The relationship between machine learning and artifical intelligence: Machine learning is one discipline contributing to the field of a.i.. It’s the learning part. There are other parts of a.i. that are not machine learning, although non-technical media treats those terms as synonyms.
- If you have received any short-and-understandable explanation on how machine learning works (I have, the best and funniest unsurprisingly being by CGP Grey), they are probably 80% wrong. That is because there are five different types of machine learning, and if you get an impression how one of them works, you are misguided when you think you understand a machine learning phenomenon that happens to use one of the other 4 types. (The 5 types according to the book being deduction, backpropagation, genetic programming, statistic inference and kernel machines. I don’t have a clue yet either.)
- Working in the field of machine learning can mean to invent new machine learning algorithms as well as applying existing algorithms to specific (business) problems. This means there are research careers and engineering careers in machine learning which are not the same.
I wanted to share these three items because I will probably forget them when I reach the end of the book – or, actually, not forget the facts but take them for granted, which means I will forget that those thoughts were new and fascinating for me and worth sharing.