Someone remarked that the sign of intelligence is the ability to reduce complex subjects into simple straight-forward explanations. This has been the best ML book I’ve encountered by that metric (and it’s only in Early Release form due in Feb 2017).
Chapter 2 is one of the best introductions to machine learning. The book as a whole covers both the breath and depth of both traditional ML and Deep Neural Networks in a style unmatched by other ML books.
All the topics are well selected, sequenced and cover topics in a uniformly in-depth, lucid and practical hands-on style. There are no weak or missing areas I could identify. Even through this is an Early Release 4 months before official publication, I found no problems in the software or demos so far.
The book primarily assumes an undergraduate course in linear algebra and some familiarity with matrix-based programming using Python Scikit-Learn. The linear algebra is essential to understand the concepts, however any matrix manipulation programming like matlab, R or Julia would make the programming implementations look very familiar.
Very highly recommended for those with Python scikit-learn and linear algebra backgrounds. Even without these, there are large parts of the book that give excellent explanations to new learners.