Is probabilistic machine learning just the mathematical background of machine learning?

Question

I wanted to begin with machine learning, I went through the contents of the course on ML by Andrew Ng and found that though the course was based on mathematics, but wasn't too much on the probability or statistics.

But in many university curriculua the book: "Probabilistic Machine Learning: An introduction" by K. Murphy is being used, and it seemed to be full of probability, statistics etc.

I wanted to know if the ML - which is being used in the industry, like the neural networks or other ML techniques - has any statistical or probabilistic foundation which is often not bothered or is it a new emerging field with mathematical basis which is not from probability or statistics?

I really want to know if the ML in academia is really the foundation part of the ML used in the industry or is it the ML in its old form?

And finally, are the contents of the book: "Probabilistic Machine Learning An Introduction", relevant in the ML being used today?

I am aware that these are quite broad and open ended questions from someone entering into the field, I would be happy to informative answers or links?

Answer 1

Knowledge in Probability and Statistics is absolutely important. It might feel you can do quite a bit without that and can get things work using the high level ML libraries that we have today. However if you are a serious practitioner you will figure out that beyond a point your knowledge on how these techniques work and are tuned is very hollow and that will automatically lead you into the more mathematically inclined books that give you the foundation knowledge.