![](https://img-blog.csdnimg.cn/20201014180756927.png?x-oss-process=image/resize,m_fixed,h_64,w_64)
机器学习
文章平均质量分 90
Kendyu
这个作者很懒,什么都没留下…
展开
-
【机器学习基础】用正则化防止过拟合|Regularization
文章目录岭回归|Ridge RegressionLasso回归|Lasso Regression岭回归 vs Lasso回归岭回归|Ridge RegressionMain idea: slightly worse fit for training data, better long term predictions. ➡️ A little more bias, significznt drop in varianceDefinition: Add λ×slopoe2\lambda\time原创 2021-02-01 18:16:25 · 104 阅读 · 0 评论 -
【机器学习算法】梯度提升方法
Comparison with AdabostGradient BoostAdaBoostInitialbuild a very short tree called stumpmake a single leaf, representing the initial guess for the target value of all samplesLoopbuild another stump based on the errors made by the previous原创 2021-01-17 16:59:50 · 239 阅读 · 0 评论 -
【机器学习算法】XGBoost
XGBoostSummary at firstXGBoost中减少计算难度的地方求叶节点的output value的时候用二阶泰勒展开求Similarity score的时候用梯度和Hessian去算RegressionFirst, an original prediction(constant) is given for all points.1. Build the treeFrom the root, calculate similarity for residual原创 2021-01-14 23:08:00 · 277 阅读 · 0 评论