Skip to content

Commit a1f8845

Browse files
committed
Add ch7-regularization-and-normalization
1 parent eaef0d0 commit a1f8845

13 files changed

Lines changed: 4778 additions & 0 deletions

_freeze/zh/.gitignore

Lines changed: 3 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -7,6 +7,9 @@ ch3-multi-layer-perceptron/
77
# Chapter 4: Optimization Algorithms
88
ch4-optimization-algorithms/
99

10+
# Chapter 7: Regularization and Normalization
11+
ch7-regularization-and-normalization/
12+
1013
# Chapter 8: Attention and Transformer
1114
ch8-attention-and-transformer/
1215

_quarto-html.yml

Lines changed: 4 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -67,6 +67,10 @@ website:
6767
contents: "zh/ch3-multi-layer-perceptron/*.qmd"
6868
- section: "Chapter 4: 优化算法:神经网络如何更新参数"
6969
contents: "zh/ch4-optimization-algorithms/*.qmd"
70+
- section: "Part 2: 卷积神经网络"
71+
contents:
72+
- section: "Chapter 7: 正则化与归一化:让深层网络更稳定"
73+
contents: "zh/ch7-regularization-and-normalization/*.qmd"
7074
- section: "Part 4: Attention 机制与 Transformer"
7175
contents:
7276
- section: "Chapter 8: Attention 与 Transformer:从动态检索到序列建模"

zh/README.md

Lines changed: 11 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -38,6 +38,17 @@
3838
- 4.8 优化器地图:不同优化算法该在什么时候使用
3939
- 4.9 学习率调度器:让学习率随训练过程变化
4040

41+
## Chapter 7: 正则化与归一化:让深层网络更稳定
42+
43+
- 7.1 为什么深层网络需要正则化与归一化
44+
- 7.2 Dropout:通过随机失活减少过拟合
45+
- 7.3 BatchNorm:利用 batch 统计量稳定训练
46+
- 7.4 LayerNorm:在单个样本内部归一化特征
47+
- 7.5 InstanceNorm:对样本的每个通道进行归一化
48+
- 7.6 GroupNorm:在通道组内归一化特征
49+
- 7.7 RMSNorm:不做均值中心化的特征尺度归一化
50+
- 7.8 归一化方法的统一视角:统计量到底在哪些维度上计算
51+
4152
## Chapter 8: Attention 与 Transformer:从动态检索到序列建模
4253

4354
- 8.1 Bahdanau Attention:从信息压缩到动态检索

zh/ch7-regularization-and-normalization/ch7.1-why-regularization-and-normalization.qmd

Lines changed: 289 additions & 0 deletions
Large diffs are not rendered by default.

0 commit comments

Comments
 (0)