动手学深度学习

“Dive into Deep Learning” is designed for learners who wish to grasp both the underlying principles and the practical implementation of models, integrating mathematical explanations, code implementations, and chapter exercises. The Chinese second edition website allows direct access to the content and provides installation instructions, notation guidelines, as well as links to courses and code resources.

The textbook first introduces data manipulation, linear algebra, calculus, automatic differentiation, and probability, before moving on to topics such as linear networks, multilayer perceptrons, convolutional networks, recurrent networks, attention mechanisms, and optimization. Learners can supplement their background knowledge according to their own level and observe model behavior through concrete tasks.

The website provides entry points for implementations such as PyTorch, NumPy/MXNet, TensorFlow, and PaddlePaddle; when making your selection, be sure to verify the corresponding section, version, and runtime environment. Online reading materials do not constitute a hosted computing service—installing dependencies, downloading data, and provisioning training compute resources all remain your responsibility.

Students, teachers, and developers who wish to systematically learn deep learning

Chinese-language textbooks are available for online reading and provide access to code, notebooks, and other learning resources; costs for the hardware or cloud computing power required to run training are billed separately.