- 向量化的好处(vectorization),简洁,高效
- linear regression
- multi feature linear regression
- feature scaling:多个特征值的取值范围相差比较大时,损失函数的等高线图会畸形,将值normalization到相差不大的范围(比如-1到1,0到1)
- 学习率:控制
- learning curve:学习曲线,随着iteration,cost J (损失函数不断减小)
- logisitic regression:二分类问题
- sigmoid function
- overfit
- how to resolve overfit
- forward propgation前向传播
second see study note
- definition of machine learning
- supervised learing (two type: regression and classsification)
- vs unsupervised learning
- 代价函数, cost function
- gradient decent 梯度下降
- gradient desccent algorithm