Reference Request: Machine Learning from Scratch

Question

I have a background in Software Engineering and the question is a bit related to CS theory, but I find it more appropriate to post my question here.

I want to get started on Machine Learning but haven't found a book that explains the process from scratch. It is either tutorials on how to use TensorFlow (but without actually learning the concepts and math involved) or more rigorous books oriented towards mathematicians (another question has been asked but with a much more rigurous approach here)

One of my favorite books is Code: The Hidden Language of Computer Hardware and Software which explains the very basics of computer science, binary and architecture in a very intuitive and approachable fashion that can be easily understood even by people with no background in the subject.

Although not that deep, I have been looking for a similar resource for Machine Learning, that walks the reader through basic calculus, statistics and linear algebra and builds up to basic machine learning. More than looking for proofs, I am interested in the math behind to aid my learning.

Answer 1

For a basic overview you could try The Hundred Page Machine Learning Book by Burkov

If you want something a bit more substantial I would suggest An Introduction to Statistical Learning by James, Witten et al. I think it is aimed at a undergraduate audience from a variety of fields. For a even more mathematical text of the same topics then try Elements of Statistical Learning by Hastie and Tibshirani, which is a classic.

Other texts that are common in the field are Pattern Recognition and Machine Learning by Bishop and Machine Learning: A Probabilistic Perspective by Murphy. Both are very readable

Answer 2

You might be interested in "Deep learning book" by Ian Goodfellow.

It starts with a bit of linear algebra and statistics(not calculus) applied to optimization problems and a general overview of machine learning followed by a deeper dive into different types of deep neural networks(feedforward, convolutional, etc.).

It's not general machine learning though, but it can be used as a starting point.