机器学习相关知识笔记(实时更新)
shrinkage and regression coefficients
https://datacadamia.com/data_mining/shrinkage
正定矩阵
https://zhuanlan.zhihu.com/p/44860862
kernel tricks
http://crsouza.com/2010/03/17/kernel-functions-for-machine-learning-applications/#kernel_trick
Quadratic function looks like
https://stats.stackexchange.com/questions/295474/how-does-a-quadratic-kernel-look-like
最小二乘法
https://www.matongxue.com/madocs/818/
协方差的引入和计算
https://www.matongxue.com/madocs/568
矩阵的rank
https://www.matongxue.com/madocs/254/
Fit和fit_transform的区别
https://zhuanlan.zhihu.com/p/42297868
矩阵的求导
https://zhuanlan.zhihu.com/p/288541909
残差
https://baike.baidu.com/item/%E6%AE%8B%E5%B7%AE
loss function & cost function
https://blog.csdn.net/UESTC_C2_403/article/details/77387780
1.损失函数(Loss function)是定义在单个训练样本上的,也就是就算一个样本的误差,比如我们想要分类,就是预测的类别和实际类别的区别,是一个样本的哦,用L表示
2.代价函数(Cost function)是定义在整个训练集上面的,也就是所有样本的误差的总和的平均,也就是损失函数的总和的平均,有没有这个平均其实不会影响最后的参数的求解结果。
梯度下降
https://zhuanlan.zhihu.com/p/33321183
kmeans
https://www.analyticsvidhya.com/blog/2019/08/comprehensive-guide-k-means-clustering/
反向传播的推导
https://www.zybuluo.com/hanbingtao/note/476663
Xgboost
https://blog.csdn.net/v_JULY_v/article/details/81410574?utm_source=app
特征值分解