Python Machine Learning By Example: Build intelligent systems using Python, TensorFlow 2, PyTorch, and scikit-learn, 3rd Edition

I think this book is worth reading if you already have a background in Machine Learning. Otherwise, a few things will be frustrating. But, I still recommend the book; just be sure you are the right audience.Let me start with the cons and then the pros.The first negative aspect of the book is that mathematical theories are not deeply covered. For example at page 18, there is the MSE which fails to explain (IMO) the "E". Similarly at page 22 when the concept of "regularization" is explained, we can argue that the whole section could be "keep it simple with the linear function" and the reader would get the same information instead of these many paragraphs. Another example that the writer miss is the set in context. Page 28 explains the One-Hot encoding but fails to justify that it improves runtime execution analysis by having boolean value. Instead, it mentions the "sparse matrix" that is well handled by "scipy" which is the first mention without much content around it. I can provide many other examples like the scaling that come into the picture way too early to see that it is worth it. No examples are provided (page 29).On the bright side, the book uses popular Python libraries. For example, it uses the Pandas, NumPy, and Matplotlib. But reading page 37, you can see that the user assume some level of expertise by assuming that we are aware of what Anaconda or Miniconda is by leaving the any explanation why the latter is preferable for casual user.Chapter two starts with a project of building a recommendation engine with the Naive Bayes. That time the author is doing a great job explaining the training data set and the prediction stage. The binary explanation is fine. However, building the recommendation engine seems to take aside with a smaller example using the MNIST with recognizing numbers and cancer examples. I found that part confusing why the main announced topic path was derailing. The Bayes theorem brings mathematical notations, which is fine, and the author ensures an example with numbers is provided to support the theory.Page 56 is where Python code starts. The quality of the representation of the code is, for me, average. The background is dark, and the type is white, which contrasts a lot. It seems that it has been copy-pasted since you can even see the "..." at each line. Not that is wrong, but some polishing could be applied. The code is commented, which is a plus. However, the code has a cons of the lack of space, making snippets extend to many pages.On page 75, the course is at recognizing a face. That is quite a feat if you never performed machine learning. But, it is motivating to exercise with an image. For some reason, I recalled from a past lecture (years ago) that SVM was not an efficient way to classify information. I do not see in this book any detail about SVM's advantages and disadvantages, which is disappointing. The chapter discusses the plane and multiplanes at the level that you are comfortable.Chapter 4 talks about a topic (IMO) that is simpler than the previous one with SVM. None the say that the subject of information gain and entropy can be confusing, but the author's writing is well suited. The only problem are the wall of code.Page 227 introduces information about linear regression, which is similar to the logistic regression previously taught. The link between them simplifies the understanding of this theory.To conclude, the book is well built. I am someone who prefer more example of a single topic, and it is not the case in the book. You get a single explanation and a single sample. Code can be hard to follow sometimes because of the length, but that is a natural disadvantage of the medium. I recommend the book for people who already have/had a background and want to refresh their skills. I am afraid that it will be too much information without the depth required to fully understand for a newcomer.

Python ML By Example (BE) is a good complement to Python ML Third Edition (3E). The 3E book focuses on the theory and general application of ML programming, while the BE book focuses an specific application examples.While they both tackle ML programming, their approach is different. The BE book assumes you have a reasonable, foundational background in ML and uses that basis to create specific ML-based applications.For example, whereas 3E has a simple note about Naïve Bayes classification, the BE book has a whole chapter dedicated to the algorithm, discussing the different types of classification methods, how Naïve Bayes works, and then actually implementing a Naïve Bayes application. On the flip side, the 3E book has a whole chapter dedicated just to the different classifiers and different implementations of them using scikit-learn.It's almost like the 3E book is a textbook and the BE book is its complementary workbook for practice. While you may be able to be successful with either one, combining them really maximizes your ML learning.To speak about the BE book in more detail, the topics covered include:*Introduction to Python ML, including software installation*Using Naïve Bayes algorithm to create movie recommendation application*Using SVM for facial recognition*Using tree-based algorithms to predict ad click-through*Using Apache Spark to work with large data sets*Using regression algorithms and neural networks to predict the stock market*Using text analysis and NLP to data mine newsgroups*Using unsupervised learning models to identify newsgroups topics*Using different types of neural networks for different types of analysis approaches*Using reinforcement learning for decision making*ML best practicesIt is a long book (nearly 500 pages), but the material is invaluable for anyone in the ML field, especially if you don't have a lot of experience with the different algorithms. And in conjunction with 3E, you almost have a complete ML curriculum.

I like the book but the typesetting of the book is very poor. I would prefer this book in color typeset

The book is a great practical resource for those interested in applying ML techniques quickly. As other reviewers have mentioned, it is a bit light on theory and the more technical aspects of ML until you get deeper into the book. Having said that, the examples and code provided are very practical and to the point. I would recommend it if you are either already familiar with basic ML theory and the math behind it or if you have a software engineering background and are simply looking to quickly implement an ML solution. The author walks you through code segments and explains each step and by the end, you will have covered most of the more popular ML models and know which ones to use given your data and your goals.The chapters on SVMs and building and predicting ad-clicks are very practical. I like how the author walks through each model type and compares them so you have a baseline and can see the differences in implementation and result. That was a helpful exercise spread across a few chapters. Also, the Best Practices chapter near the end was a good, "I need an answer quickly" kind of reference.Overall, I liked the book and think it would be helpful from a more practical perspective. I think this book paired with another more theoretical resource would really round out those seeking to learn ML.

I'm looking forward to the latest and updated edition which covers the rise GPT 4.0 and Foundation models.This book is a nonsense , straightforward approach.Yes in some parts its a little outdated however , if you debug and correct it you can get it working.This is a book with code in it , the game changes quickly.

Has all the things you need to know.

Though I liked the book as it takes a balanced approach to a topic by including concepts and code, I was disappointed on reaching page 7790. The next page is 7821... 30 pages are missing from the Kindle version that I purchased

The concepts are explained clearly and step by step. Machine learning is not an easy topic, I find that this book is really helpful