En mi caso me interesó este libro pues la mayoría del tiempo trabajo con R y he estado poco a poco aprendiendo python, con principal intereses en el aprendizaje profundo. No he tenido la oportunidad de trabajar con keras o tensorFlow y tengo que decir que me sentí algo intimidado cuando mire los ejemplos que se encuentran que la web del desarrollador. Este libro en apariencia es extenso con sus 495 páginas y 9 capítulos pero se debe a que tiene varias ilustraciones, bloques de código junto con el detalle de las funciones. Desde el primer capítulo hasta el último se definen y aclaran conceptos de manera muy sencilla sin entrar en detalles matemáticos, si ya tienes cierta experiencia podrías saltarte el capítulo 1. Lo genial del libro es que hace una perfecta sincronía entre mostrar los pasos que se deben seguir para la elaboración de un modelo y cómo usar las bibliotecas keras y tensorflow. Además el libro está estructurado de tal manera que todo el desarrollo y aprendizaje de un capitulo se usa para el capítulo siguiente.Todo el código que se muestra en el libro ya se encuentra en github junto con el archivo requirements.txt que facilita mucho la instalación de las bibliotecas necesarias. Recomiendo mucho este libro a personas que tengan poca o nada de experiencia en el desarrollo de modelos ya que hará que su primer contacto sea muy fácil y muy enriquecedor en conceptos y buenas prácticas al momento de crear un modelo de aprendizaje profundo. No lo recomendaría para personas con experiencia que buscan el concepto matemático o que desean profundizar en la optimización o uso de los hyperparametros de un modelo.In my case I was interested in this book as most of the time I work with R and have been slowly learning python, with main interests in deep learning. I have not had the opportunity to work with keras or tensorFlow and I have to say that I felt a bit intimidated when I looked at the examples found on the developer's website. This book in appearance is lengthy with its 495 pages and 9 chapters but that is because it has several illustrations, code blocks along with the detail of the functions. From the first chapter to the last it defines and clarifies concepts in a very simple way without going into mathematical details, if you already have some experience you could skip chapter 1. The great thing about the book is that it makes a perfect synchrony between showing the steps that must be followed for the development of a model and how to use the keras and tensorflow libraries. Also the book is structured in such a way that all the development and learning from one chapter is used for the next chapter.All the code shown in the book is already on github along with the requirements.txt file which makes it much easier to install the necessary libraries. I highly recommend this book to people with little or no experience in model development as it will make your first contact very easy and very enriching in concepts and best practices when creating a deep learning model. I would not recommend it for experienced people who are looking for the mathematical concept or who want to go deeper into the optimization or use of the hyperparameters of a model.

One of the best books on Keras!

This is a great book if you want to learn how to use Keras for deep learning! The layout is clear and easily understandable, with lots of practical applications with code examples. The book is well organized into chapters that make it a practical resource. Practical aspects such as data preprocessing, model evaluation and advantages/disadvantages of deep learning are also covered, which make this a good reference for both how deep learning models work as well as how they can be best applied. Highly recommended for anyone looking to start using Keras for their machine learning projects!

The authors do not assume any prior knowledge from the reader except for some experience with Python and Linear Algebra. However, the chapters manage to avoid lengthy introductory theory and cover the necessary basics quickly which means that you’ll see how to actually use the packages on example problems in no time. The book relies heavily on examples, which keeps the pace high, and offers explanations on what is going on behind the scenes along the way.All in all, I think it is an excellent resource for people new to the field.Best read with a Jupyter notebook running on aside.(FYI, I was sent a copy of the book by the authors for an unbiased review )

I’ve always found neural networks and deep learning fascinating. The ease with which they can be used to solve any machine learning problem are mind-blowing, though their computational power varies with the number of neurons you decide to use, which can be either a hindrance or a boon to the model.The advantage of having open-source platforms, such as Python, Jupyter, and the several libraries and modules available is there is always a software developer who has already tried to solve the problem you are working on. Thus, the likelihood that there is a Python library capable of implementing machine learning or deep learning models was already pretty high. Enter Tensorflow and Keras.Both these libraries work together, and they are very easy to implement straight out of the box – once you’ve installed the necessary dependencies that enable their use. And though the tutorials on their website are straightforward, there is very little documentation on how to apply neural networks beyond what is offered on the website, which is simple regression and image classification problems.The Deep Learning with Keras Workshop aims to correct this.It is a very easy to follow guide which strives to lay the groundwork behind how neural networks work and how these could be applied using Keras. It does so by assuming that the user isn’t wanting in statistical analysis or how modeling works, but by first explaining the proper way the data should be assessed, establishing a baseline using both the data itself and a simple model, and from there building up on how neural networks work.When it comes to how models are built, evaluated, and tuned, it is very easy to find tutorials that cover all this; albeit these can be very brief, not explaining the intricacies of the model, while also being very crowded when it comes to explaining how each part of the model-building process comes together. I thoroughly enjoyed that, in this book, while each of these parts is studied, they are each dedicated a separate chapter, allowing the reader to take their time to explained how concepts such as hyperparameter tuning or cross-validation are used to boost the model’s prediction capabilities. By not cramming model building, tuning, evaluation, and cross-validation into every single chapter, but rather presenting each one individually and using the previous as a foundation to build on the next, the authors have done an extremely good job building a book that genuinely teaches not only how Keras works, but how the process of building any model works.Turning to the user examples, each chapter builds on the previous one, allowing the user to build more robust models, but not all chapters use the same examples. I found it refreshing that the book works with several real-world problems and data sets, showing the user how neural networks and Keras can be applied to several problems. And, though not all of them have high accuracy rates or low-test errors, which is not a downside given this is how most real-world problems work, it helps the user by showing the applications of neural networks beyond simple regressions and image classification.The one true downside to this book is that it does not go into detail how each parameter affects the model. Why is it important to have more (or less) epochs or batches? What is the difference between an “adam” and “sgd” optimizer? What is the difference between using Dense(1, activation=’sigmoid’) and Activation(‘sigmoid’)? Nevertheless, going into these details could make the book harder to understand and could cause the user to lose interest, so perhaps adding some minor details, or pointing to further material that can go in depth into those subjects, akin to what is done with the Keras and Tensorflow package websites, would be helpful.Overall, I find that my experience with Keras and neural networks has improved upon reading this book, and though the use of such is computationally demanding, I find that I can now try to use small deep learning models as part of my model evaluation and comparison when exploring new data sets in a more comfortable manner.