Dive into Deep Learning — 书籍拆解

读到哪:未读。 readState 不是 read/partial 的书不能当锚。

项
作者	Aston Zhang、Zachary C. Lipton、Mu Li、Alexander J. Smola
版次	持续更新(GitHub master @23d7a5ae,2026-08-23 取)
格式	md ｜ 文本源 raw
许可	CC BY-SA 4.0
来源	GitHub 仓库 d2l-ai/d2l-en @23d7a5ae markdown 源(作者团队在 GitHub 以 CC BY-SA 4.0 开放发布,2026-08-23 取)
清洗	删页眉页脚 0 行、页码 0 行、断词接回 0 处

我们重写的拆解(0 章)

（还没写。拆解是这本书对我们的真正产出——底下的元数据只是索引。）

为什么收它

深度学习的开放教材,markdown 源、章节结构天然完整,是本书架里转码质量最好的一本。 讲义里要解释模型机制时可以直接锚到具体小节。

合法性

作者们以 CC BY-SA 4.0 在 GitHub 开放发布;上游 LICENSE 原文存在 library/dive-into-deep-learning/LICENSE.upstream。原始文本不入库。

它大概覆盖什么、不覆盖什么

覆盖:从线性回归到注意力/Transformer 的完整脉络,带可运行代码。 不覆盖:agent 架构、产品工程——那些在三个代码书架里。

它覆盖什么、不覆盖什么

（还没读到能下判断的程度。claims / notCovered 空着就是空着,不猜。)

怎么引用它

(依据: book=dive-into-deep-learning §Introduction)

章节名对不上会被 lab:validate 拦下;页码锚(§p.123)同样可用。

结构(168 段,共 3003k 字符)

段	章节	页	规模
01	Preface	—	26.1k
02	Installation	—	6.1k
03	Notation	—	4.3k
04	Introduction	—	83.9k
05	Data Manipulation	—	22.6k
06	Data Preprocessing	—	7.2k
07	Linear Algebra	—	33.0k
08	Calculus	—	15.4k
09	Automatic Differentiation	—	18.0k
10	Probability and Statistics	—	43.9k
11	Documentation	—	4.6k
12	Linear Regression	—	32.3k
13	Object-Oriented Design for Implementation	—	21.4k
14	Synthetic Regression Data	—	11.7k
15	Linear Regression Implementation from Scratch	—	21.4k
16	Concise Implementation of Linear Regression	—	15.5k
17	Generalization	—	19.0k
18	Weight Decay	—	18.8k
19	Softmax Regression	—	25.3k
20	The Image Classification Dataset	—	12.3k
21	The Base Classification Model	—	9.4k
22	Softmax Regression Implementation from Scratch	—	13.4k
23	Concise Implementation of Softmax Regression	—	10.1k
24	Generalization in Classification	—	23.8k
25	Environment and Distribution Shift	—	33.0k
26	Multilayer Perceptrons	—	23.9k
27	Implementation of Multilayer Perceptrons	—	11.8k
28	Forward Propagation, Backward Propagation, and Computational Graphs	—	12.0k
29	Numerical Stability and Initialization	—	15.2k
30	Generalization in Deep Learning	—	16.7k
31	Dropout	—	17.8k
32	Predicting House Prices on Kaggle	—	17.6k
33	Layers and Modules	—	25.5k
34	Parameter Management	—	9.6k
35	Parameter Initialization	—	11.6k
36	Lazy Initialization	—	7.5k
37	Custom Layers	—	9.2k
38	File I/O	—	8.4k
39	GPUs	—	18.7k
40	From Fully Connected Layers to Convolutions	—	17.7k
41	Convolutions for Images	—	22.6k
42	Padding and Stride	—	15.8k
43	Multiple Input and Multiple Output Channels	—	14.2k
44	Pooling	—	14.9k
45	Convolutional Neural Networks (LeNet)	—	15.3k
46	Deep Convolutional Neural Networks (AlexNet)	—	28.8k
47	Networks Using Blocks (VGG)	—	14.2k
48	Network in Network (NiN)	—	12.0k
49	Multi-Branch Networks (GoogLeNet)	—	24.3k
50	Batch Normalization	—	44.7k
51	Residual Networks (ResNet) and ResNeXt	—	36.2k
52	Densely Connected Networks (DenseNet)	—	19.5k
53	Designing Convolution Network Architectures	—	25.3k
54	Working with Sequences	—	24.8k
55	Converting Raw Text into Sequence Data	—	12.5k
56	Language Models	—	16.3k
57	Recurrent Neural Networks	—	13.3k
58	Recurrent Neural Network Implementation from Scratch	—	26.7k
59	Concise Implementation of Recurrent Neural Networks	—	7.6k
60	Backpropagation Through Time	—	19.5k
61	Long Short-Term Memory (LSTM)	—	21.6k
62	Gated Recurrent Units (GRU)	—	15.3k
63	Deep Recurrent Neural Networks	—	12.8k
64	Bidirectional Recurrent Neural Networks	—	8.5k
65	Machine Translation and the Dataset	—	12.7k
66	The Encoder--Decoder Architecture	—	8.6k
67	Sequence-to-Sequence Learning for Machine Translation	—	34.5k
68	Beam Search	—	10.4k
69	Queries, Keys, and Values	—	11.2k
70	Attention Pooling by Similarity	—	12.8k
71	Attention Scoring Functions	—	29.9k
72	The Bahdanau Attention Mechanism	—	23.4k
73	Multi-Head Attention	—	18.4k
74	Self-Attention and Positional Encoding	—	18.4k
75	The Transformer Architecture	—	52.6k
76	Transformers for Vision	—	16.5k
77	Large-Scale Pretraining with Transformers	—	26.3k
78	Optimization and Deep Learning	—	11.4k
79	Convexity	—	21.0k
80	Gradient Descent	—	20.3k
81	Stochastic Gradient Descent	—	18.1k
82	Minibatch Stochastic Gradient Descent	—	28.9k
83	Momentum	—	19.2k
84	Adagrad	—	15.9k
85	RMSProp	—	7.8k
86	Adadelta	—	6.5k
87	Adam	—	12.3k
88	Learning Rate Scheduling	—	24.1k
89	Compilers and Interpreters	—	20.6k
90	Asynchronous Computation	—	12.4k
91	Automatic Parallelism	—	9.5k
92	Hardware	—	34.8k
93	Training on Multiple GPUs	—	20.6k
94	Concise Implementation for Multiple GPUs	—	12.1k
95	Parameter Servers	—	13.7k
96	Image Augmentation	—	18.2k
97	Fine-Tuning	—	15.6k
98	Object Detection and Bounding Boxes	—	5.9k
99	Anchor Boxes	—	42.8k
100	Multiscale Object Detection	—	9.0k
101	The Object Detection Dataset	—	8.8k
102	Single Shot Multibox Detection	—	31.9k
103	Region-based CNNs (R-CNNs)	—	12.5k
104	Semantic Segmentation and the Dataset	—	17.3k
105	Transposed Convolution	—	10.9k
106	Fully Convolutional Networks	—	15.6k
107	Neural Style Transfer	—	21.2k
108	Image Classification (CIFAR-10) on Kaggle	—	22.7k
109	Dog Breed Identification (ImageNet Dogs) on Kaggle	—	20.4k
110	Word Embedding (word2vec)	—	12.6k
111	Approximate Training	—	7.8k
112	The Dataset for Pretraining Word Embeddings	—	15.7k
113	Pretraining word2vec	—	11.7k
114	Word Embedding with Global Vectors (GloVe)	—	10.3k
115	Subword Embedding	—	10.0k
116	Word Similarity and Analogy	—	8.3k
117	Bidirectional Encoder Representations from Transformers (BERT)	—	27.2k
118	The Dataset for Pretraining BERT	—	18.9k
119	Pretraining BERT	—	13.1k
120	Sentiment Analysis and the Dataset	—	7.4k
121	Sentiment Analysis: Using Recurrent Neural Networks	—	9.7k
122	Sentiment Analysis: Using Convolutional Neural Networks	—	15.3k
123	Natural Language Inference and the Dataset	—	11.8k
124	Natural Language Inference: Using Attention	—	19.2k
125	Fine-Tuning BERT for Sequence-Level and Token-Level Applications	—	9.7k
126	Natural Language Inference: Fine-Tuning BERT	—	16.7k
127	Markov Decision Process (MDP)	—	7.2k
128	Value Iteration	—	14.7k
129	Q-Learning	—	13.4k
130	Introduction to Gaussian Processes	—	17.9k
131	Gaussian Process Priors	—	15.6k
132	Gaussian Process Inference	—	30.3k
133	What Is Hyperparameter Optimization?	—	15.4k
134	Hyperparameter Optimization API	—	13.2k
135	Asynchronous Random Search	—	10.9k
136	Multi-Fidelity Hyperparameter Optimization	—	12.2k
137	Asynchronous Successive Halving	—	10.0k
138	Generative Adversarial Networks	—	19.9k
139	Deep Convolutional Generative Adversarial Networks	—	24.4k
140	Overview of Recommender Systems	—	6.2k
141	The MovieLens Dataset	—	9.3k
142	Matrix Factorization	—	10.7k
143	AutoRec: Rating Prediction with Autoencoders	—	7.4k
144	Personalized Ranking for Recommender Systems	—	6.8k
145	Neural Collaborative Filtering for Personalized Ranking	—	14.2k
146	Sequence-Aware Recommender Systems	—	13.2k
147	Feature-Rich Recommender Systems	—	7.1k
148	Factorization Machines	—	7.6k
149	Deep Factorization Machines	—	7.0k
150	Geometry and Linear Algebraic Operations	—	39.6k
151	Eigendecompositions	—	22.2k
152	Single Variable Calculus	—	26.1k
153	Multivariable Calculus	—	43.7k
154	Integral Calculus	—	21.7k
155	Random Variables	—	44.0k
156	Maximum Likelihood	—	15.0k
157	Distributions	—	27.5k
158	Naive Bayes	—	20.1k
159	Statistics	—	33.0k
160	Information Theory	—	40.9k
161	Using Jupyter Notebooks	—	6.1k
162	Using Amazon SageMaker	—	5.4k
163	Using AWS EC2 Instances	—	10.7k
164	Using Google Colab	—	1.3k
165	Selecting Servers and GPUs	—	9.8k
166	Contributing to This Book	—	8.2k
167	Utility Functions and Classes	—	44.9k
168	The d2l API Document	—	2.9k

我们自己的读书笔记(0 篇)

（还没有。读完某章后写进 docs/dive-into-deep-learning/notes/,那才是这本书对我们的产出。）

---

本页由 node scripts/book-build.mjs 生成:表格来自转码结果,散文来自书卡正文。不要手改本页。