Inside Deep Learning: Math, Algorithms, Models is a book I have been working on for the past year since 2019 and has grown out of the course material for my deep learning class at UMBC. The idea is I wanted to write a book that would teach someone all of the practical skills and techniques within Deep Learning that I find end up being useful on-the-job. In particular I want readers to build an intuitive understanding of why and how Deep Learning works, something deeper than "this is a toole / API" but not as intimidating as a common graduate textbook. Targeting the "middle ground" between these two extremes that has been an underserved area. It was also critical to me that the book be accessible without buying a GPU. Too many smart people simply dont have the funds to invest $600+ into compute resources to find out if they maybe do or do not like deep learning. For this reason I've designed every example in the book to run in under 15 minutes using a free Google Colab GPU instance, with most running in under 5 minutes. This way you do not need to invest up-front to learn. I'm very proud of my students who have worked through this material, and been able to read and even implement recent peer reviewed papers after completion. I hope you will find the same utility and success from this book!
The table of contents are listed below, you may note that the order of topics is somewhat unusual compared to most materials. I've intentionally striven to place the topics in an order where each chapter can grow upon the previous ones in a way that builds a understanding and appreciation for why the techniques exist. For example, Transfer learning is covered far later than most books - but in my experience my students have walked away with a much deeper appreciation for why, when, and how to use transfer learning.
- The Mechanics of Learning
- Fully Connected Networks
- Convolutional Neural Networks
- Recurrent Neural Networks
- Modern Training Techniques
- Common Design Building Blocks
- Auto Encoding and Self Supervision
- Object Detection
- Generative Adversarial Networks
- Attention Mechanisms
- Network Design Alternatives to RNNs
- Transfer Learning
- Advanced Building Blocks
