Review of Learning Deep Learning: Theory and Practice of Neural Networks, Computer Vision, Natural Language Processing, and Transformers Using TensorFlow, by Magnus Ekman (2021)
- Donāt have to be a researcher to read the book. Itās perfectly fine for a sharp Python engineer.
- Book examples are written with TensorFlow, but online Jupyter Notebooks are available for both PyTorch and TensorFlow with permissive MIT license
- While I would have preferred to see PyTorch in the printed book, I need to do the hands-on PyTorch examples anyhow, so it might be even better that Iām forced to use the Jupyter Notebooks.
- Book explains the differences between TensorFlow and PyTorch.
- Comprehensive coverage of DL techniques that are industry standards for the past 2-3 years, but have rarely been collected in one place in book form until this book. See Hands-On Machine Learning with Scikit-Learn, Keras, and TensorFlow for another book that encyclopedically covers DL tricks and techniques, also using TensorFlow (but w/o PyTorch examples)
- Pointers to research papers throughout.
-
Nice charts at the end with the algorithms and ātricks of the tradeā for both vision and natural language processing applications. Some tricks work for some, not all applications. Engineers usually work in one field, not both - and have a difficult time sorting out what techniques are specific to their field.
- While reading this book, I also read Deep Learning with PyTorch, and I follow Stanfordās CS231n course: Convolutional Neural Networks for Visual Recognition (Spring 2017), both freely available online.
- Deep Learning with PyTorch has its own Jupyter notebooks (check out my own edited version of the notebooks).