吴恩达深度学习2笔记week1——深度学习的实用层面 Setting up your ML application

1.1 训练_开发_测试集 train/dev/test sets

1.2 偏差_方差 Bias/Variance

1.3 机器学习基础 Basic “recipe” for machine learning

1.4 L2正则化 Regularization

1.5 为什么正则化可以减少过拟合？Why regularization reduces overfitting?

1.6 Dropout 正则化 Dropout regularization

• Inverted dropout 反向随机失活
  
  

1.7 理解 Dropout Understanding dropout

• dropout是一种正则化方法，有助于预防过拟合
  

1.8 其他正则化方法 Other regularization methods

• 数据扩增 data augmentation
  
• early stopping
  

1.9 归一化输入 Normalizing inputs

1.10 梯度消失与梯度爆炸 Vanishing/exploding gradients

1.11 神经网络的权重初始化 Weight initialization for deep networks

• weight 与1接近
  

1.12 梯度的数值逼近 Numerical approximation of gradients

1.13 梯度检验 Gradient Checking

• help find bugs in implemntations of back propagation
  
  

1.14 关于梯度检验实现的注记 Gradient Checking implementation notes