Machine Learning with TensorFlow

This book is a fantastic hands-on introduction to machine learning. Tensorflow is widely used and a prominent player in the machine learning library space.The book is well written, and the code is available on github. I would recommend this book to any software engineer or student trying to get their feet wet with machine learning.It clearly explained seq2seq models in a way I've never seen before. This is better explained than most online tutorials. And other books can dive too far into formulations that lose sight of the intuition, but this book explains the material in a timeless manner.

Wishing to learn about TensorFlow, I decided to survey TF books available from Amazon, and pick one or two for further study. I excluded self-published offerings, and ended up with this longish list, dominated by Packt titles:"Machine Learning with TensorFlow" by Shukla, published by Manning in 2018-02, 272 pp, $43"Mastering TensorFlow 1.x" by Fandango, Packt, 2018-01, 474 pp, $35"Pro Deep Learning with TensorFlow" by Pattanayak, Apress, 2017-12, 398 pp, $37"TensorFlow 1.x Deep Learning Cookbook" by Gulli and Kapoor, Packt, 2017-12, 536 pp, $32"Neural Network Programming with TensorFlow" by Ghotra and Dua, Packt, 2017-11, 274 pp, $40"Predictive Analytics with TensorFlow" by Karim, Packt, 2017-11, 522 pp, $50"Machine Learning with TensorFlow 1.x" by Hua and Azeem, Packt, 2017-11, 304 pp, $39"Learning TensorFlow" by Hope and Resheff, O'Reilly, 2017-08, 242 pp, $25"Hands-On Deep Learning with TensorFlow" by Van Boxel, Packt, 2017-07, 174 pp, $35"Deep Learning with TensorFlow" by Zaccone, Karim, Menshawy, Packt, 2017-04, 320 pp, $50"TensorFlow Machine Learning Cookbook" by McClure, Packt, 2017-02, 370 pp, $30"Building Machine Learning Projects with TensorFlow" by Bonnin, Packt, 2016-11, 291 pp, $35"Getting Started with TensorFlow" by Zaccone, Packt, 2016-07, 180 pp, $35I reviewed the doc on tensorflow.org - including the doc for older releases - then started looking at books. Books except Shukla's, I should say - this one I ordered in hard-copy, and it took a while to arrive.By the time it did, the picture became reasonably clear. The available choices split into (a) titles that I would not recommend, (b) titles that could be somebody's best choice, depending on one's preferences. The first group included the books by Zaccone, Karim, Zaccone and Karim, Bonnin, Hua and Azeem, Ghotra and Dua, and (at full price) Van Boxel. The second group included the books by Pattanayak (a surprisingly rigorous almost-textbook of the ML methods associated with TensorFlow, but not really a TensorFlow learning aid), Hope and Resheff (an introduction that starts off well and then goes into the weeds; I gave the book five stars, but now want to reduce that rating), two similar books by (i) Fandango and (ii) Gulli and Kapoor (less polish and depth, more coverage) - I will call this cluster FGK - and, finally, the book by McClure, with slightly more depth but a lot less "width" than FGK, so sort of sitting between FGK and Hope-Resheff. In this group, I gravitated towards FGK and picked Fandango.Enter "Machine Learning with TensorFlow". Will it crush the Packt quickies? Will it be the book I settle on? I open the book and read its first lines. "Learning to parallel park a car for the first time is typically an intimidating challenge. The first few days are spent getting familiar with the buttons, assisting cameras, and engine sensitivity". Buttons? What buttons? How about mirrors? As I spend the next five minutes learning about Park Assist, the fog of confusion clears, but I am left wondering why neither the author nor the editor thought twice about this sentence. The particular sin of picking an unfortunate example will be committed again and again - for example, the first regression on display will lack the intercept, and in the section on unsupervised learning, the author will choose an example from audio processing, a niche interest - but the feeling that competent editors and reviewers were not there will be with me constantly. Did anyone notice that the book's definition of a "model" on page 18 defines a *deterministic* model, which is kind of odd for a book about statistical learning? (No wonder the author then has trouble explaining under/overfitting). Did anyone, on the other hand, take a course on Markov chains, and advise the author that, contrary to his repeated claims, history can be incorporated into a Markovian model, by simply including x(t-1), x(t-2), etc. in the state vector? The chapter on HMMs, praised by another reviewer, starts off well but follows it up with cryptic code supposedly related to the Viterbi algorithm - is that really good enough? The chapter on reinforcement learning is a complete write-off, failing to explain RL or the presented (dodgy) exercise - you look for how that model defines the state, and you find it, sort of, in comments to code, and it is not what the code itself says.I accept that the book's (a) accessible (even chatty) writing style, (b) chapter-length introduction to ML in Chapter 1, and (c) serviceable first chapters will make it an appealing choice for many TensorFlow novices. My advice would be to not look for an "all-in-one" package, and get statistics/ML education from a dedicated book - "Elements of Statistical Learning" by Hastie and Tibshirani and "Introduction to Statistical Learning" by James et al. come to mind - and rely on TensorFlow books just for TensorFlow.("Machine Learning with TensorFlow" goes back to Amazon, and I am taking a closer look at Fandango).

Very well organized book with great examples.

Would of given it a 3.5 if had the option. Machine learning with tensor flow is good introductory text for those with some python experience. It covers all the basics from regression/classification to CNN/LSTM and chatbots. Gives a good look at the skeleton of the framework. The cons - Due to the constantly changing TF versions, including the addition of new abstractions, it seems like it’s slightly outdated already. It also would have been nice to have a chapter exhibiting tensor flow’s distributed computing capabilities using google cloud ML one of the key advantages to using TF.

I do not recommend buying this book. It is very basic, it does not have mathematical foundations, and everything that is explained is at a very low level. It is informative, and not formative, and everything I could read in it, I found it much better on the internet, and in several articles. I think buying it was lost money. It is not worth the money invested in the few pages it has, and in the poor content.No recomiendo comprar este libro. Es muy basico, no tiene fundamentos matematicos, y todo lo que se explica, es a un nivel muy bajo. Es informativo , y no formativo, y todo lo que pude leer en el , lo encontre mucho mejor en internet , y en articulos varios. Creo que comprarlo fue dinero perdido. No vale el dinero invertido en las pocas paginas que tiene, y en el pobre contendio.

This book took way too long to publish (I followed along as a Manning Early Access Program member), and the examples don't seem to work and use the outdated/deprecated APIs. I know it's hard to keep up with the rapid pace of changes to Tensorflow, but it feels like the author gave up at some point and someone else dusted off the transcript and shipped it as is.

Really poor, circuitous writing. The tensorflow documentation is both more accurate and more concise. Broken examples.

I thought I'd have a hard time reading this, but this book made everything easy to follow. It offered tons of examples with annotated code.

Very simple introduction to TensorFlow, touching different aspects of machine learning, without going into much detail (and sometimes this is a bad thing, as it is not always clear, why certain operations are performed. The main shortcoming of the book is that the presented code, especially towards the end of the book, becomes incomplete or faulty. Fortunately, the book contains links to the code online, and there you can find a working version.

I really enjoy reading this book. Well written, good information for almost all levels.

I’m a fan of Manning Books as they are usually quite helpful. However, in this case I’m disappointed. Half through the book now, I noticed that there are many errors in the code snippets shown in the book. This means that I need to resort to online sources to resolve code issue. This is exactly NOT what I want to do when I buy a book that is supposed to provide good guidance.The book partly focuses on “simple” things such as linear regression, logistic regression etc, instead of sticking to the topic of the book (which is the application of tensorflow with all the details).Overall I would not buy the book again and I cannot recommend it.

This well written book is very useful both for beginners in ML and for experienced ML but with no exposure to tensor flow.The Editorial policy of "pay once - deliver many" is really good: I purchased the paper edition since that is what I prefer for traveling and also for home reading, however the online version with copy paste and the mobile version kindle and mobile devices are delivered "all inclusive" and enable very efficient learning.

Das Buch ist nicht besonders dick. Ich konnte es innerhalb von drei Stunden leicht durchlesen. Wenn man die Beispiele nachprogrammiert dürfte man etwas länger brauchen, wobei die Beispiele recht leicht/einfach sind.Etwas Fortgeschrittene finden nichts neues darin, aber als Anfängerbuch zum Einstieg in das Themengebiet voll zu empfehlen.