[체험판] Large Scale Machine Learning with Python

▶Customer Review
⦁Python + Machine Learning + Scalability Under One Cover
: Many books when their subject is Machine Learning with Python concentrate on a few most known and used libraries to explain Machine Learning tasks and solutions. Although I don't want to say that such books are useless for readers, they may still leave gaps in understanding of how a certain method or library would work in real-world scenarios. Authors of the book "Large Scale Machine Learning with Python" set up an ambitious goal to teach readers how to solve real-world Machine Learning problems by employing a variety of libraries, frameworks, and tools relying on Python. This advantageously differentiates a given book from many other books on the same subject.

The following practical situations are considered and their solutions are presented:

- Tall datasets when the number of cases is large, compared to the number of features.
- Wide datasets when the number of features is large, compared to the number of cases.
- Both tall and wide datasets when both the number of features and the number of cases are large.
- Sparse datasets when there are many zero-valued elements.

The book treats the problem of scalability from different angles, such as fast batch (offline) processing, incremental online processing (one instance at a time arrives), streaming processing (a chunk of instances at a time arrives) and distributed processing. Popular libraries and frameworks, such as Gensim, H2O, XGBoost, TensorFlow, Theano, Theanets, Keras, Vowpal Wabbit, and Spark and their applications are explained through numerous Python snippets. In my opinion, this is one of the first books presenting all these tools under one cover.

In addition to Python code, the book also covers such advanced topics like Deep Learning, Ensemble Learning, validation of streaming algorithm performance, and GPU processing.

I recommend this book as a good companion to any Machine Learning practitioner who already has fairly good understanding of theory behind Machine Learning algorithms.
(-by Oleg Okun)

⦁Excellent data science resource, even if you are not an adept in Python
: This is the best book for Python-based data science, focusing on ML and big data I have encountered (and I’ve been around!). The authors cover a wide-range of intermediate and advanced topics, which they explain in terms of theory and applications. I particularly liked the Unsupervised Learning chapter, where they not only covered the quite popular k-means algorithm, but also provided a couple of heuristics for finding the optimum number of clusters while they wrote a few words about one of its most powerful variants (k-means++) too.

Although Python falls short when it comes to handling large data sets or multiple CPUs/GPUs on its own, the authors describe the various solutions to these issues via the use of large scale frameworks, such as Spark, making Python a versatile tool for big data scenarios. Also, they introduce the various packages required to accomplish all the analytics-related tasks, making this book also a great reference manual for all data scientists who veer towards this language.

Personally I lean towards more elegant and more modern programming tools, such a s Julia and Scala, but I found this book quite refreshing and insightful, definitely a great addition to my data science library. If you are someone who takes data science seriously and has learned the basics, I would highly recommend this book for you.
(-by Z.V. )