Deep Learning with TensorFlow and Keras: Build and deploy supervised, unsupervised, deep, and reinforcement learning models, 3rd Edition

If you prefer to learn things with hands-on coding then this book is for you. It is expected that you have Tensorflow with high level Keras API already setup (or simply use google Colab), some exposure to deep learning basics along with some experience Python as well.In the first chapter book provides a brief intro to Keras and Perceptron, look at various activation functions and jumps to first code example. This chapter covers a lot of ground such as regularisation, gradient decent, learning rate etc with practical MNIST example. Book cover a number of deep learning topics such as natural language processing, Auto Encoder, GAN, CNN, reinforcement learning etc. each section includes good code examples.Extra points for very well laid out structure of all chapters. I know many people who wants to learn ML/DL but they often get put-off by the mathematics behind it. Book cleverly introduce maths behind machine learning in a separate chapter in later part of the book. Before reaching to this chapter reader already get familiarised with magical Keras and Tensorflow libraries which make building ml model so easy with only 40-50 line of codes and great flexibility. This helps reader to delve into behind the scene mathematics with some confidence and practical knowledge obtained through rest of the chapter.Book includes several references to the source materials (i.e. research papers, earlier work) for each type of neural nets, this allow you to gain further and deeper knowledge if like to get into the details.I have ML background mostly with CNN, GAN and some exposure to RNN. Graph Neural Nets was very interesting chapter for me.There are plenty of practical examples to learn from which is quite amazing.

[Some context:- I was kindly provided with a digital review copy in exchange for my feedback.- I run a Machine Learning Practical university course and am approaching this as a potential supplementary book to recommend to my students.- This is a quick review, based on my impressions of the material covered. I have not verified the correctness of all material in the book, nor have I run the code in it].The book has very good coverage in terms of Machine Learning topics relevant to Deep Learning. This largely explains its size, at near 700 pages. Each area then has its most salient subtopics covered, though not at any significant depth. A lot of the space is dedicated to fully print out example code.From what I have sampled, the text is well written and understandable, with good use of examples and figures to accompany the points made. However, it is worth noting that these are somewhat rushed through, in the sense that points are not deliberated on, and a lot of subtopics are only served with a quick introduction. A necessity perhaps given the book's breadth.If one can follow the text, and is happy to supplement it with further material in order to clarify or reinforce the occasional point, then the layout of the subtopics within each chapter will provide a comprehensive overview of its corresponding topic. This is made easier by a good selection of reference material, both in the form of code/libraries and research.I would advise any potential buyer to go through a sample of Chapter 1 (available at least on Amazon). This chapter is a good example of the book's approach. If you can follow it without too many leftover questions, or you find it easy to locate further material where you need a clarification, then the book will prove useful to you. If you find it hard to do so, then you will either need a book that goes slower (and deeper) on the fundamental theory (which could supplement this one) and the code explanations (unfortunately the available sample does not make use of much code but what is there is representative of their approach). Note, that the book's Discord page is active and can be useful in this endeavour.I expect the book would serve also as a good quick-start for someone who needs to learn hands-on, provided they have a reasonable background in programming, linear algebra, and calculus (or you might not be clear on some operations).Overall, this is a very good reference book and starting point for a broad range of applications. I am likely to use select chapters as recommended starting points for students on relevant projects (quick access to baseline solutions and a quick overview of relevant ideas and resources). After trying it out this way, select chapters might be useful for integration into the theory part of my course. I would warn anyone who aims to be an engineer or analyst to supplement this book with a book/course delving deeper into, especially, the fundamentals. As long as you don't uncritically follow the examples in the book (for example, Chapter 1 might leave you thinking it is a good idea to repeatedly test on your held out test data; it is not) you will find it a valuable reference and starting point for any given project.

As we all know that in the AIML domain next to Machine Learning the Deep Learning specific projects are highly complex, more challenging and in high-demand areas. Personally, I have come across many books and YouTube videos, but I was looking for just one sort of block to cover my expectations, full fill the demands and for those who are all looking for Deep Learning perspectives along with TF and Keras framework, I got that feeling while review and read the book called “Deep Learning with TensorFlow and Keras” by three authors.Here I would like to share my observations based on my interesting chapters base from this book.Based on my experience in the AIML space, there is a very close relationship between mathematics and data science, machine learning and deep learning, so believe the authors have properly picked up and evidently summarized this in Chapter 14 – “The Math Behind Deep Learning”. I would say this topic is very valuable to understand the various mathematics behind Deep Learning concepts.The authors have covered many things and the interesting notions I come across on the list are - Gradient descent, Activation functions of Derivative of sigmoid, tanh, ReLU, and backpropagation – Forward and Back Steps are quite interesting and most important before we apply them in our model building.The objective of the book is DL with TensorFlow, I can suggest that readers should go through thisChapter 19: TensorFlow 2 Ecosystem, to understand better before we use them with other major topics like RNN, CNN and other topics. Here we can explore various subcomponents under TF 2 ecosystem as TF Hub, Datasets, Lite and its architecture along with different edges and applications, Federated learning, and Node.js with TensorFlow models.Chapter 15: Tensor Processing Unit, is a major topic and backbone of the whole book since this TPU is a special chip developed at Google for ultra-fast execution of neural network mathematical operations additional improvement with TPUs is to remove from the chip any hardware support for graphics operations normally present in GPUs.The one-line takeaway from this topic is “TPU is a special purpose co-processor specialized for deep learning, being focused on tensor specialized operations. The authors have discussed all four generations of TPUs and Edge TPU, really this is a special gift for me and of course readers as well.In Chapter 2: Regression and Classification, the Authors have given the feel of Regression, Classification models and their types and how to use TF, and Keras API with these algorithms with classical examples, please try these samples and understand how TF and Keras work with these basic algorithms.Chapters 3 & 20: CNN and Advanced CNN – Authors are kick-starting how Convolutional Neural Networks (CNN) leverage spatial information, architectures, and other elements which highly recommended while doing DL experiments, and they are sharing how CNN is well-suited for classifying image-specific problem statements. And on top, they are providing the overview of CNNs, DCNNs, and ConvNets along with the sample code step by step using CIFAR-10 images. The topics about the very deep convolutional networks for large-scale image recognition understanding and deep inception V3 are compelling content for readers.Followed by this they were offering a detailed outline of how CNNs can be applied within the areas of computer vision, video, textual documents, audio, and music in Chapter 20. Here we could understand how to use CNN's for text processing and computer vision principles. Under image segmentation, we would extract the U-Net architecture, and Fast/Faster/Mask R-CNN network architecture.Guys, please read “A summary of convolution operations” this is must read topic to understand different convolution operations and how I x O x K parameters are used in CNN.Chapter 6: Transformers – This chapter super power pack for DL engineers, authors have clearly described the transformer-based architectures very precisely. Particularly in the NLP space. Authors have detailed Categories of transformers and popular transformers like BERT, RoBERTa, ALBERT, StructBERT, DeBERTa,GTP2 & 3, Reformer, Transformer-XL, XLNet the list is long-lost and excellent implementation is truly credited to readers to understand on one book itself as a collective note.Chapters 4, 5 & 17:The authors have focused on text data in quite an interesting way, starting with word embeddings and various techniques and methods used in the industry like – Static, Neural, Character and subword, Dynamic, Sentence and paragraph and Language-based model Recurrent Neural Networks, Authors have presented an exceptional feast for DL lovers in slice and dice [cells, cells variants, topologies] form and the core role of RNN in various problem-solving areas such as speech recognition, language modeling, machine translation, sentiment analysis, and image captioning, and LSTM and GRU topics are again must-read.Authors have covered the relatively new class of neural networks is nothing but the Graph Neural Network (GNN), As we know that is ideally suited for processing graph data. In the current demand, many real-life problems in areas such as social media, biochemistry, academic literature, and many others are inherently “graph-shaped,” meaning that their inputs are composed of data that can best be represented as graphs, personally this special gift for me. They have coved Graph basics, Graph machine learning, Graph convolutions and Common graph layers and their applications and way to customizations.I would say this book surely will teach you the complete Deep Learning family and understand the concepts in the best way.All the best to the authors. Overall, I can give 4.0/5.0 for this. Certainly, an extraordinary effort from the authors is much appreciated.-Shanthababu PandianArtificial Intelligence and Analytics | Cloud Data and ML Architect | Scrum MasterNational and International Speaker | Blogger

Nearly 700 pages of indispensable references and examples on how to use two of the most popular and AI and ML libraries out there. I have just skimmed through it currently and was literally blown away at how many examples this book provides. To me having examples to follow when you are new to learning this stuff is essential and a must and this book does not lack in that department.As others have already mentioned it gives a great introduction on the subject and talks about what TensorFlow and Keras is and what these libraries are built for. It gets in depth into things like Regression and Classification models. CNN's, DCNN's, RNN's, GAN's, and a ton more!!!It has a dedicated chapter on just the math behind Deep Learning which is really amazing for all the big math buffs out there. I like it because it is not hard to follow and you can easily follow along to most of the book with very little math background which is why having libraries like TF and Keras is so powerful. It gives programmers a simple to use API that allows anyone with the desire to start working on their own projects the tools to do so without being bogged down by the difficult things which have been abstracted away to such a level that you can now run models with less then 100 lines of code. So amazing when you really think about it.This book not only provides easy to follow examples, and writing style, it has a ton of references for each chapter giving the reader the opportunity to follow up with anything they want to dig deeper into which is very powerful for finding additional information on the subjects. This is especially useful for students and researchers who are actively writing peer reviewed papers on this subject.I would recommend this book to anyone who is interested in learning about or anyone who may already be involved in AI and ML, or just anyone in general. You can know nothing and this book will be helpful to you or you can already be a seasoned expert and this book will still be valuable to you, it is just a well rounded and jam packed book full of useful examples and knowledge.I am very happy I picked this book up.

Concepts are clearly explained. Should have machine learning knowledge as prerequisite.

One of the best resources to learn the tricks of the trade of building cutting edge Machine and Deep Learning systems. Authors team has put in efforts to clearly explain the fundamentals of ML and DL. Extensive code samples are helpful for successful learning.This book provides just right theory and practice required to use Keras, TensorFlow, and AutoML to build machine learning systems....

The book is great in the sense that it covers a wide array of machine learning methods and provides a lot of depth to most of them. The downside, though, is that the book is littered with typos.Most of the typos have to do with equations or with model results. For example, the book incorrectly gives the equation for a plain vanilla LSTM and instead gives the equation for a peephole LSTM. And as far as the results go, they often show the Python output of a model and then describe the results using different outcomes than the ones presented in the results.The typos do not make the book unreadable. But they do require you to double check model equations at every turn to make sure the authors got it right (they don’t enough times for it to matter).For a third edition, these typos are unexpected and obnoxious. I’d still recommend getting the book, with the caveat that the typos make it necessary to read closely and double check using resources aside from this book.

The media could not be loaded.