d2l en

Written by Mu Li, Alex Smola, and colleagues from Amazon and academia, the book uniquely integrates mathematical theory, conceptual explanation, and executable code examples in a unified formatevery concept is illustrated with code that readers can run, modify, and experiment with, producing immediate feedback that builds intuition alongside formal understanding.

The textbook covers the complete arc of modern deep learning: linear neural networks, multilayer perceptrons, model selection and regularization, convolutional neural networks and computer vision, recurrent networks and sequence modeling, attention mechanisms and transformers, optimization algorithms, computational performance, NLP applications, and generative adversarial networks.

Code examples are provided in multiple frameworks (PyTorch, TensorFlow, JAX) so readers can apply the material within their preferred ecosystem. The open-access format makes the full text available freely online.

Students in university deep learning courses, self-directed learners building structured ML expertise, and working practitioners seeking a rigorous reference that combines mathematical depth with practical implementation use d2l.ai as a primary or supplementary text.

The book's approach of co-presenting theory and runnable code has influenced how subsequent ML educational materials are structured, and its breadth of coveragefrom fundamentals to advanced topicsmakes it useful across the full spectrum of ML competency levels from undergraduate students to experienced researchers encountering unfamiliar topics.