TensorFlow For Machine Intelligence: A hands-on introduction to learning algorithms

by Sam Abrahams

This book is a hands-on introduction to learning algorithms. It is for people who may know a little machine learning (or not) and who may have heard about TensorFlow, but found the documentation too daunting to approach. The learning curve is gentle and you always have some code to illustrate the math step-by-step.

TensorFlow, a popular library for machine learning, embraces the innovation and community-engagement of open source, but has the support, guidance, and stability of a large corporation. Because of its multitude of strengths, TensorFlow is appropriate for individuals and businesses ranging from startups to companies as large as, well, Google. TensorFlow is currently being used for natural language processing, artificial intelligence, computer vision, and predictive analytics.

TensorFlow, open sourced to the public by Google in November 2015, was made to be flexible, efficient, extensible, and portable. Computers of any shape and size can run it, from smartphones all the way up to huge computing clusters. This book starts with the absolute basics of TensorFlow. We found that most tutorials on TensorFlow start by attempting to teach both machine learning concepts and TensorFlow terminology at the same time. Here we first make sure you've had the opportunity to become comfortable with TensorFlow's mechanics and core API before covering machine learning concepts.

Genres: Technology

342 pages, Kindle Edition

Published July 23, 2016

Reviews

[Vincent Russo · Rating 3 out of 5]

It seems like this book was really rushed. When I see a decent amount of spelling mistakes in a book, it really makes me think that the rest of the book's quality is equally lacking. Examples:

"We use it squared to avoid computing the square root..." -- We use it squared?
"As a quick evaluation, you can check that the model learned how the blood fat decays with weight..." -- The blood fat content?
"...and sigma is the greek letter..." -- Greek is a proper noun.

I could go on, but I won't.

Indeed, the coverage and layout are just plain clunky. Even the quality of the code presented in the book is subpar. For instance, the line ``saver = tf.train.Saver()'' has the comment "Create a saver". How useful is that comment? It's not, it's worthless. Spelling the word "through" as "thru", etc. This isn't even the lot of it, there's just messy commenting throughout the code, and the explanation in the books could be done in a much more in-depth. Another downfall of this book is the way that the math is rendered. To the author's credit, I read this book on the Safari website, so it's possible the math formatting is an artifact of the proprietary reader that Safari provides. The equations are in significantly larger font for some reason, and in certain equations involving matrices, the entries are so obscured it's impossible to tell what the entries are.

Now this book isn't all bad, there's a decent amount to be gleaned from it in terms of what you can do with Tensorflow, and perhaps I should give more credit. The amount of books on Tensorflow is, at this time, still fairly few and far between. So a book that goes into detail about what you can do with actual examples is pretty useful. I think it would be worthwhile to release a follow-up edition, but the way that the book is currently written just seems relatively slapped together.

[Xuankang Lin · Rating 3 out of 5]

I read the official documents first. It's not read-friendly there so I skimmed through this book.

For me, this book only adds several things helpful in addition, such as using jupyter notebook... It's unrealistic to fully understand CNN/RNN by this book. Moreover, this book is for API v1.0, there are a bunch of API changes already.

I would recommend using Stanford CS20si course as a start instead.

[Jonas · Rating 4 out of 5]

I really like the book, and would recommend it if you are looking for a beginners introduction to TensorFlow. Few others that i've read on TensorFlow are basically just a compilation of tutorials and docs from the web. In the first half, this book does indeed a good job explaining the principles on which TensorFlow was built, provides nice example how to structure your code and some really accessible machine learning examples. I would give 5 stars for the first half of the book. The second half where the book talks about more advanced topics seems a bit rushed. Where are quite a few spelling mistakes, examples are not self contained, you'll need to fill some parts. Nevertheless, I liked the parts on CNNs and RNNs. The chapter on TensorFlow serving could have a bit more details.

[Yilmaz Aksoy · Rating 4 out of 5]

20 yıl sonra yapay zekaya yeniden giriş yaparım ve TensorFlow öğrenirim diye okudum bu kitabı. Ne yazık ki örnek programları çalıştırmakta çok zorlandım. Oldukça yeni bir kütüphane (2015) için yazılmış bu kitap tabii ki kütüphanedeki arabirimlerin değişmesinin kurbanı olmuş. Yapay sinir ağları örneklerinde programları artık zorla da çalıştıramayınca sadece okumakla yetindim. Daha önce yapay sinir ağları programlamış olduğumdan anlatılanı anladım ama herkes için böyle olacağını sanmıyorum.

Hızlı bir giriş kitabı ve girişte kalıyor, modern teknikler ile ilgili iki bölüm var. Kesinlikle bu konuda daha başka kaynaklara da başvurmam gerekecek. Python bilmek kitabı anlamak için daha da faydalı olacaktır, en azından konuya odaklanmak daha kolay olacak, yoksa benim gibi ikide bir başka yerlerden "bu ne demek acaba?" diye araştırma yapmak gerekebilir.

[Karan · Rating 4 out of 5]

It's a good book to polish your TensorFlow concepts. It systematically reveals the 'thought' with which TensorFlow models should be made in the initial chapters. Later chapters are about examples and implementation where the author has skipped some of the details (and made some spelling errors!).

As others have also mentioned, some of the code is outdated, thanks to TensorFlow developers for such speedy developments with a lot of deprecations and little backward compatibility.

Overall, a good guide to understanding the workflow of making your TensorFlow models, but not to copy the code and write models.

[Nick · Rating 1 out of 5]

Very disappointed. Errors in the source code and using deprecated TensorFlow functions prevented me from finishing this book. Installation instructions for Jupyter Notebook is wrong; it should be installed into the created TensorFlow virtual environment, not in system-wide (sudo).

[Mark · Rating 2 out of 5]

This book is well and clearly written but unfortunately the subject matter has changed so much since publication that most of the examples won’t even run, and since the content relies heavily on those examples it’s a significant setback. A new edition would be very valuable.

[Ferhat Elmas · Rating 4 out of 5]

Partially outdated and I recommend official docs but it might be a good starter to learn concepts before diving docs that contain many Tensorflow specific terms.

[Alex · Rating 2 out of 5]

One of the better introductory books on Tensorflow, which could've been a lot better if the errors were fixed and the contents brought up to date with non-deprecated TF API.
I cannot deny that narrowing down alternatives to given code in current APi is not a worthy learning experience in itself, but this is time that could've been spent better elsewhere.