Learn TensorFlow 2.0: Implement Machine Learning and Deep Learning Models with Python

by Pramod Singh

Learn TensorFlow 2.0
Chapter 1: TensorFlow 2.0 - An Introduction Chapter Introducing TensorFlow, major features, version 2.0 release.
Chapter 2: Supervised Learning with TensorFlow 2.0 Chapter Implementation of linear, logistic, SVM (Support Vector Machines) and random forest using TensorFlow.
Chapter 3: Neural Networks and Deep Learning with TensorFlow 2.0 Chapter Introduction to neural networks, deep learning and implementation using TensorFlow This chapter offers a detailed view of building Deep Learning models for various applications such as Forecasting using TensorFlow 2.0. The chapter also introduces optimization approaches and the techniques for hyper parameter tuning.
Chapter 4: Images with TensorFlow 2.0 Chapter TensorFlow 2.0 for images. This chapter focuses on building deep learning based models for image classification using TensorFlow 2.0. It covers advanced techniques such as GANs and transfer learning to image processing and classifications
Chapter 5: Sequence to Sequence Modeling with TensorFlow 2.0
Chapter To understand sequence modeling using TensorFlow 2.0. This chapter covers the process of using different neural networks for NLP based tasks in TensorFlow 2.0. This includes sequence to sequence prediction, text translation using deep learning in TensorFlow 2.0
Chapter 6: TensorFlow 2.0 Models in Productionization
Chapter Implementation of distributed training using TensorFlow. This chapter covers the process of scaling up the machine learning model training by implementing distributed training of TensorFlow models and deploying those models into production using TensorFlow serving layer

182 pages, Paperback

Published December 21, 2019

Reviews

[Jin Shusong · Rating 2 out of 5]

Any one who is familiar with Keras should not read this book since the authors just introduce simple tf.keras apis. There are no contents about tensorflow 2.0 low level apis.

[Jason Braatz · Rating 5 out of 5]

This is a 5-star book and well worth the read (as of December, 2020) for those who are interested in getting up and running on some basic and intermediate training models with a recent version of Tensorflow.

I read this book at the same time as Introduction to Machine Learning, as I am in the early phases of learning the subject. This book (compared to the other one just mentioned) is on implementation, not theory, so for a reader who doesn't quite know why multiple convolutions are needed or when to prune a decision tree when over-fitting is happening will not get the gist. There has to be a minimum level of understanding on ML/Deep concepts first before jumping into one implementation of an ML library (this one being Tensorflow).

Bottom line: this is a great book for those who are having to make an infrastructure or implementation decision. Python is a slow scripting language and while Tensorflow also works with R, there are very few use cases (I can imagine) where this library makes sense in an outwardly facing application. The framework is too young, for one, and the extensibility is just beginning to emerge. As another reviewer wrote, this book does not go into the new low-level API that Tensorflow has now exposed to the development community. I think this was on purpose: this was a book written for someone being introduced (perhaps for the first time) to Tensorflow.

This is actually a very good book for a CEO, CTO, Data Scientist or even an IT Manager. This is a rapidly changing field so it's hard to get ones arms around it exhaustively for any period of time. Luckily, the TF API is about as easy as it gets, aside from web services like Peltarion AI which does a great job in making a widely diverse topic achievable for us mere mortals.

It's also a much lighter read than most books within the AI field. At the moment, we have more mathematicians and physics maestro's in this field than we have software developers. Honestly, I can't help but think the ratio is 2:1 between the number of Data Scientists (and alike) versus the number of developers fully versed on TF and TF-Lite implementations. So I applaud the author in doing a great job of writing a very developer-centric book.

I'm skeptical about the longevity of TensorFlow itself: unless you are a 3D artist, you aren't likely developing on a machine with dual nVidia 3080 GPUs, so it makes a lot more sense to use Google's co-lab tool (which lets you experiment a bit in a Jupyter notebook without having to forklift your hardware, since it's using Google's TPUs - not whatever one that's local on the machine). More and more we're seeing developers opt for portability (Chromebook, MacBook Air, etc) over power, and it simply makes sense given the competition to buy higher end TPUs/GPUs -- it's been a price war with the coin mining crowd and youtube gamers over these cards and so I can't imagine that a local or in house solution (like a locally deployed TensorFlow) will end up ever being anything useful for large datasets. That said, the author predicted this and do a very nice job of bringing in examples of doing experiments on a free Google colab account.

5 Stars - well worth the read for those who are breaking into ML/AI, and would like focus on experimentation over commercial app implementation .

[Arash Amani]

I read it in less than 10 days. It gives a good sense when I have learned TF 2.0. when I was trying to learn TF 1.0 and it was not comfortable for me. This book gives you an overlook of Kears too.
TF 2.0 is easier to learn than TF 1.0.