Deep Learning: A Practitioner's Approach 1st Edition

This is an excellent book. I have hundreds of papers and books on Neural Nets from the time of Rosenblatt's Perceptron on through autoencoders, recurrent NNs, convolutional NNs, RBM's, DNN's, greedy pretraining, Kolmogrov's universal approximation theorem, optimization methods for weight training, and more.I found this book to provide a conceptual overview of the DNNs and the architectures (feed forward, deep belief, unsupervised pre-trained, convolutional, recurrent, long and short term memory, and recursive, networks). The book provides the conceptual connective tissue that are the muscles that the practitioner must bond to the architectural bones to move forward in Deep Learning. The book is a remarkable debrief by two lead developers of the DL4J framework; Josh Patterson and Adam Gibson. Every chapter offers new nuggets about how to apply their framework to real world ML problems, and about real world ML problems.The ending chapters are about the actual application of the DL4j framework to practical problems, and how to use the framework with DL4j with Spark, the ND4J API, using GPU's, distributed training, and trouble shooting.Unlike most books, many of which I buy for reference purposes to find what I need at various times, I am reading this one page-by-page to pick up all of the insightful observations. The book is an easy read for practitioners, and well worth the time, and modest price.DL4J may provide some real competition to Tensorflow, and Caffe, especially in enterprise Java environments.Hats off to Josh Patterson and Adam Gibson. Well done.

Have you been following the Artificial Intelligence in computing fad for the last decade or so? I have since I designed an AI based medical device to regulate human blood pressure in the ICU and received FDA approval for my design to be used in the US. I wanted to know what others were doing in applying AI in various computing categories and which ones were similar to natural neural networks in mammalian brains. But the literature was too scattered and apparently incomparable to make sense in general to be useful.Well, this is the book we've been looking for and it's about time! This is the gateway book to almost all of the methodologies used in developing AI computing. I still uniquely own the knowledge of developing AI by expert system design. But, in 500 pages this book covers the introduction to deep learning, fundamentals, architectures, concepts and models, tuning, data vectorization, and Spark data reduction with Hadoop. I found more areas of AI being uncovered here than I knew existed. What a bonanza!Designers are all much richer now that we can incorporate these AI approaches into our thinking. Buy the book and become an AI expert overnight. There is just one caveat, you will have to buy additional references to get to the deep details of the learning process in each category. But at least, you will have the relative certainly of knowing that you have examined all of the known approaches and picked the one most appropriate to be successful for your application.

I am hoping this book earns five stars, and will come back and update this rating if the typo I found on day one is an aberration. I ordered this book back in May and was very pleased to get it; so far it is excellent and exactly at the level I need: I am a sometime practitioner of machine learning and AI using a range of open source and off-the-shelf tools. I have moved more strongly into Python as i mostly deal with text and NLTK and python have bee easier / faster for me to pick up and use than java-syntric approaches. So moving into deep learning is a big step but I feel I am well prepared, and the level and degree of "refreshers" here, from linear algebra tp statistics are hitting just the right depth and tone.I initially thought my Kindle software was broken when searching for the first occurrence of "SGD" didn't show up; I remembered it was referred to as the "canonical" solution to solving a system of linear in an iterative fashion, but forgot what it stood for. Sure enough, right there on page 15 you find "The canonical example of iterative methods most commonly seen in machine learning today is Stochastic Gradient Descent (SDG), which we discuss later in this chapter.". I spent far too longing cross checking my paper book, my kindle reader (mac software) and notes before "seeing" the "SDG" was not "SGD"... so i will give a full five stars if this is a rare occurrence. Having source code on GitHub at least means I am less worried about such issues with code samples...

I agree with the other three-star reviewers: the fundamental first three chapters are poorly organized and seem hastily written. Concepts jump from very basic to very advanced, some with little more than a mention (ex: just 3 sentences on clustering), so the reader does not get "a full journey through deep learning", at least not a useful one. The title "Deep Learning" is misleading: it should really be something like "A Practical Guide to Deep Learning with DL4J".

The book is based on the powerful open source deeplearning4j framework, which aims mainly for Java and other JVM languages (e.g. Scala).Deeplearning4j is an efficient and easy to use system and the book uncovers its potential very well. The book avoids the rather difficult theoretical discussions, and instead provides the necessary intuition for applications in real problems.The book has strong focus on the application of deep learning models, and it presents clearly and in easy to understand way a lot of applications.It is an excellent book, that can be used effectively with the more theoretical "Deep Learning" book of Ian Goodfellow, Yoshua Bengio, Aaron Courville, in order to gain both theoretical and applied insight on the emerging field of deep learning.In a few words, it is a superb book, especially for Java/Scala programmers.