A Very Brief Review of Some Books on Deep Learning
In an effort to understand and deepen my understanding of Deep Learning (DL), one of the hottest topics in AI, over the past 3-4 years, I've purchased several books on the topic on Amazon. I'm sharing brief reviews of seven of them below, hoping they'll be helpful to you and your colleagues. Coincidentally, I'm teaching a course on Deep Learning this semester.
- Deep Learning, Goodfellow dkk., MIT Press, 2016
This book's discussion of deep learning is very fundamental and covers the fundamental mathematical concepts necessary for a good understanding of deep learning. It's an excellent book on deep learning theory and serves as a reference for many scientific works in the field. For readers without a strong mathematical understanding, reading some sections of this book can be daunting. DL topics covered include Deep Neural Networks, regularization techniques, optimization algorithms, Convolutional Neural Networks, sequence modeling, etc., with applications in computer vision, speech recognition, natural language processing, recommendation systems, etc. As a book focused on fundamental concepts, there are no coding examples included for various deep learning concepts and algorithms. Therefore, for hands-on experience with DL, readers need to complement it with one of the following books, such as No. 2, 3, 5, or 6. The new edition is eagerly awaited by many readers, including myself, of course. It's especially recommended for academics and practitioners who want to understand the fundamentals of deep learning, complete with a number of slides.
- Hands-on Machine Learning with ScikitLearn and Tensorflow, Aurellien Gerron, O'Reilly, 2017(first edition), 2019
This book discusses machine learning & deep learning with a discussion that balances mathematical concepts. and application examples and coding. Coding is provided in ScikitLearn, which is a Python machine learning library, and Tensor Flow. Hands-on experience is emphasized in this book. The discussion of the mathematical aspects is not as heavy as the book by Goodfellow. This book is divided into two major parts. The first part discusses the fundamentals of machine learning. The second part discusses neural networks and deep learning. The print quality of the first edition of this book (2017) was not good, especially the mathematical formulas, but it seems to have been improved in the 2019 edition. This book is quite thick. Recommended for both academics and practitioners.
- Deep Learning with Python, Francois Chollet, Manning, 2017
book written by the creator of Keras. This book is very readable, providing a fairly comprehensive perspective on machine learning and deep learning, with many coding examples in Python. Discussing ANN and DL at the architectural level, without going into the details of the learning algorithms, but good enough for readers to understand the basic concepts of DL and utilize them for various applications. This book discusses machine learning, neural networks, deep learning, and their applications in computer vision, text processing, and time series. There is a special chapter on generative models, with the topic of auto-encoders and an introduction to generative adversarial networks (GAN). Coding examples are given in Python, and the front-end uses Keras. Recommended for practitioners and academics.
- Neural Networks and Deep Learning, Charu Aggarwal, Springer, 2018
book about ANN and DL with fundamental & mathematical concepts and algorithms discussed in depth like Goodfellow's Deeplearning book. The discussion of this book is divided into three major parts: 1. Basics of Neural Networks, 2. Fundamentals of Neural Networks, 3. Advanced Topics in Neural Networks (Deep Learning). Some advanced topics such as Deep Reinforcement Learning and GAN are discussed at the end of Part 3. There are no coding examples in this book. Therefore, for a hands-on experience, readers need to complete it with one of the books such as No. 2, 3, 5 or 6. Recommended for academics, and is quite good to use as a textbook for DL courses at the postgraduate level, equipped with slides.
- Introduction to Deep Learning, E. Charniack, MIT Press, 2019
This book introduces Deep Learning by guiding readers to understand Deep Learning from the basics and gradually moving on to more advanced topics using Python and Tensor Flow. Each chapter is equipped with programming examples. This book discusses ANN, Convolutional NN, Recurrent NN, Deep Reinforcement Learning, Unsupervised Models, etc. Each chapter of this book is equipped with programming projects, exercises, and a bibliography. Readers are expected to have an understanding of linear algebra, calculus, probability, and statistics. This book is good for beginners, academics, and practitioners, and is easy to read.
- TensorFlow for Deep Learning : From Linear Regression to Reinforcement Learning, Ramsundar & Zadeh, O’Reilly, 2018
This book discusses Deep Learning using TensorFlow. The architectures and algorithms presented include linear and logistic regression, deep neural networks, hyperparameter optimization, convolutional neural networks, recurrent neural networks, reinforcement learning, and large deep networks. The book is not thick, only 240 pages, making it easy to read, just like Charniack's book. Although not thick, this book presents essential DL architectures and algorithms for understanding DL. Recommended for academics and practitioners.
- Deep Reinforcement Learning Hands-on, Maxim Lapan, Pakt, 2018 (edisi terbaru 2020)
To solve problems involving decision-making such as in games and robotics, reinforcement learning (RL) algorithms can be used effectively. This book discusses RL-based methods with an emphasis on advanced RL topics that integrate RL with deep networks. It is equipped with coding examples and their applications, especially in games. Readers' knowledge of DL and Python (and Pytorch) basics is expected to be able to understand Deep RL well. This book can be seen as a complement to the classic RL book by Sutton and Barto which discusses more about RL theory, while this book focuses more on implementation. I only have the first edition (2018). The latest edition seems to add some advanced topics such as multi-agent systems & RL in robotics.
Pengulas : Bambang Riyanto, STEI-ITB, Center for AI-ITB, www.riset.ai