Boek review Pattern Recognition and Machine Learning

Pattern Recognition and Machine Learning is often the first book I consult when I need to know something about a data subject. It is written by Christopher Bishop in a clear, smooth style. It is completely “Bayesian”, but Bishop also deals with a large number of classical methods.

The book covers, among other things, Bayesian regression, model selection, classification, neural networks, kernel machines, PCA, MCMC, etc. With over 700 pages it seems an intimidating book, but it is worth your time. You will need to refresh your knowledge about vector calculus, but then you will experience how complete the theory is covered.

Pros: Bishop’s mathematical notation – I have not seen a better notation anywhere and therefore use it in my Bayes in Action course. The examples are didactically clear, well thought out, and are particularly beautiful.

Cons: Some parts of the theory are just explained in the exercises, of which only part of the solutions have been made available. Not all data used in the examples were made available, in particular the “oil flow data”. I never understood Chapter 8 on “Graphical Models”, but that was not an obstacle to the rest of the book.

The book is very suitable for self study. I learned a lot from it. The book deserves the highest score from me ★★★★★.