Deep learning. Immersion in the world of neural networks (AI column)

I finally got my hands on reading this book, which had been on my shelf for five years:) This book came out. 2018 Since then, a lot of water has flowed into the world of deep learning, but the book is still interesting. I liked it because it was written by Russian authors who are well-managed with humor and mathematical foundations and have no problems with translation:) The book has three parts and ten chapters, the titles of which make it clear how interesting the book is. (Each part, chapter and subchapter are accompanied by funny epigraphs.)

Part part 1. How to train neural networks Head. 1. From Biology to Computer Science, We Need to Go Deeper Head. 2. Preliminary information, or the course of the young fighter Head. 3. The Perceptron, or Embryo of the Wise Computer Part part 2. Basic architecture Head. 4. Faster, Deeper, Stronger, or On Ravines, Valleys, and Swamps Head. 5. Convolutional Neural Networks and Autocoders, or Don’t Believe Your Eyes Head. 6. Recurrent neural networks, or how to bite your tail Part part 3. New architectures and applications Head. 7. How to teach a computer to read or Mathematician - Man + Woman Head. 8. Contemporary Architecture, or How Truth Is Born in Dispute Head. 9. Deep Learning with Reinforcement, or the Amazing Champion Accident Head. 10. Neurobayev methods, or the past and future of machine learning

In general, the book looks at pretty much everything that happened in deep learning. 2018 year, including references to the brain, theorver and matstats, gradient descent, convolutional and recurrent neural networks, distributed word representations (word2vec)Attention and encoder-decoder models, deep reinforcement learning and neurobaesian techniques. To read a book, it is very useful to be savvy in mathematics at the level of a university course - it will allow you to better capture the magic that happens inside models and understand why it works this way, rather than just messing with the code in Tensorflow or Keras.

P.S. I'd love to read an updated reissue of the book looking at what's happened in recent years. 5+ years, especially if the authors retained their theoretical and practical presentation and humor.

#AI #Math #Software #ML #Humor