R中套索、弹性网和岭回归的不同惩罚函数

Question

在 R 中使用正则化（如套索、脊或弹性网）时，它们是否允许我将惩罚函数更改为 Huber 或绝对值，而不是二次 L2 范数？

Answer 1

I think there is no package, as of right now, that allows to specify your own formula for the penalization. There are special packages like hqreg or ncvreg that give you some other options. The first one hqreg allows for Huber. The second ncvreg allows for minimax concave penalty (MCP) and smoothly clipped absolute deviation (SCAD).