[체험판] Bayesian Analysis with Python

▶Customer Review
⦁ Nice intro to Bayesian data analysis
: This is one of those books that attempts to fill the gap between dry theory and practice. Bayesian approach to data analysis has been around for long time. Although intuitive explanation of the different topics can be found here and there in form of tutorials, YouTube videos, a practitioner also needs examples that he/she can understand and run.
The book is very accessible in my opinion as long one has some basic Python skills. The first few chapters introduce the whole paradigm and the way of thinking for data analysis using Bayesian approach.
- Chapter 1 will get Bayesian concepts covered;
- Chapter 2 will introduce the PyMC3 python package (diagnosis is very valuable part and perhaps needs to be beefed up);
- Chapter 3 introduces hierarchical models that often are a good fit for real-world data (make sure you understand the concept of shrinking!)
- Chapter 4-5 deals with linear models for regression and classification; one will very likely use one of those in practice
- Chapter 6 is about model comparison and nicely shows the concepts of overfitting and underfitting

I didn't read completely the Ch. 7-8. The Ch. 8 on Gaussian processes is an exciting one and has many applications to modeling complex real-world phenomena. This part is likely to be outdated as in PyMC3 this module (?) has seen a lot of changes/updates. See the questions in discourse and Bill Engels' tutorials (very informative).

Overall, the book is pretty balanced with all necessary concepts introduced and with many examples. On the downside, I would revisit some chapters with large blobs of text and factorize them in more manageable chunks. It is better to write less but be more precise. This is what a 2nd edition can rectify I guess.
(-Vladislavs Dovgalecs)

⦁ Excellent introductory book on Bayesian Statistics in Python
: It is not easy to find materials for a short introductory course in Bayesian Statistics, especially if you want to use PyMC3, and this book gives you all that. The book is highly practical, and goes much more in-depth than "Bayesian Methods for Hackers" or "Think Bayes". All the codes are in Jupyter notebook on Github so that the students can follow quite easily even without much Python experience. It does not assume too much knowledge in Probability or Statistics, and the pages on this is a bit limit. However, there is a reading list at the end of each chapter for the motivated student. All in all, it is a great book to kick start Bayesian Statistics.
(-Charles Fribourg)