main idea:用adaboost类似的方法，选出g，然后选出步长
![[11-3] Gradient Boosting regression](https://image.miaokee.com:8440/aHR0cDovL2ltYWdlcy5jbml0YmxvZy5jb20vYmxvZy82MTAyMzkvMjAxNTAyLzAyMjI0NTM0OTUzMzA3MS5wbmc%3D.png?w=700&webp=1 "[11-3] Gradient Boosting regression")
Gredient Boosting for regression：
![[11-3] Gradient Boosting regression](https://image.miaokee.com:8440/aHR0cDovL2ltYWdlcy5jbml0YmxvZy5jb20vYmxvZy82MTAyMzkvMjAxNTAyLzAyMjI0NTQzMzQzNjAzNC5wbmc%3D.png?w=700&webp=1 "[11-3] Gradient Boosting regression")
![[11-3] Gradient Boosting regression](https://image.miaokee.com:8440/aHR0cDovL2ltYWdlcy5jbml0YmxvZy5jb20vYmxvZy82MTAyMzkvMjAxNTAyLzAyMjI0NTUwMDc4NTkyNS5wbmc%3D.png?w=700&webp=1 "[11-3] Gradient Boosting regression")
h控制方向，eta控制步长，需要对h的大小进行限制
![[11-3] Gradient Boosting regression](https://image.miaokee.com:8440/aHR0cDovL2ltYWdlcy5jbml0YmxvZy5jb20vYmxvZy82MTAyMzkvMjAxNTAyLzAyMjI0NTU4ODI4MzkzMi5wbmc%3D.png?w=700&webp=1 "[11-3] Gradient Boosting regression")
对(x,残差)解regression，得到h
![[11-3] Gradient Boosting regression](https://image.miaokee.com:8440/aHR0cDovL2ltYWdlcy5jbml0YmxvZy5jb20vYmxvZy82MTAyMzkvMjAxNTAyLzAyMjI0NjAxODEyODk3NS5wbmc%3D.png?w=700&webp=1 "[11-3] Gradient Boosting regression")
对(g(x),残差)解regression，得到eta
![[11-3] Gradient Boosting regression](https://image.miaokee.com:8440/aHR0cDovL2ltYWdlcy5jbml0YmxvZy5jb20vYmxvZy82MTAyMzkvMjAxNTAyLzAyMjI0NjE1MDYyNzc1Ny5wbmc%3D.png?w=700&webp=1 "[11-3] Gradient Boosting regression")
相关文章
- How to Configure the Gradient Boosting Algorithm
- 【机器学习】笔记--梯度提升（Gradient boosting)
- Linear regression with multiple variables(多特征的线型回归)算法实例_梯度下降解法(Gradient DesentMulti)以及正规方程解法(Normal Equation)
- Gradient Boosted Regression
- 回归树，Gradient Boosting和GBDT
- flink 批量梯度下降算法线性回归参数求解（Linear Regression with BGD(batch gradient descent) ）
- 斯坦福机器学习视频笔记 Week1 线性回归和梯度下降 Linear Regression and Gradient Descent
- scikit-learn : GBR (Gradient boosting regression)
- 机器学习中的数学(1)-回归(regression)、梯度下降(gradient descent)
- 斯坦福机器学习视频笔记 Week1 Linear Regression and Gradient Descent