Dive into Deep Learning 动手学深度学习

Dive into Deep Learning (D2L) is an open-source textbook that teaches deep learning through math, figures and runnable code side by side. Each section is a Jupyter notebook you can run and modify.

What's inside

The book goes from the basics (linear algebra, calculus, probability, automatic differentiation) through linear models, multilayer perceptrons, CNNs (LeNet, AlexNet, VGG, ResNet, DenseNet), RNNs, LSTMs and GRUs, attention and Transformers, optimisation algorithms, computer vision, natural language processing, recommender systems, GANs, reinforcement learning and Gaussian processes. It includes practical chapters on GPUs and multi-GPU training, and Kaggle-style exercises.

Details

• Authors: Aston Zhang, Zachary C. Lipton, Mu Li and Alexander J. Smola. The print edition is published by Cambridge University Press.
• Frameworks: implementations in PyTorch, NumPy/MXNet, JAX and TensorFlow.
• Adoption: the site says it has been adopted at 500 universities in 70 countries.
• Chinese edition: 《动手学深度学习》, also free online and a bestseller in print.
• Free to read online; the source is on GitHub.

Good for

Students and engineers who want a hands-on course rather than a purely theoretical book. Pair it with courses on Coursera or practice on Kaggle.