Linear penalized regs

[anzonyquispe]

Translate the r-code from this notebook to python script.

[alexanderquispe]

1. If the l1_ratio corresponds to alpha , why we have different values for alpha and l1_ratio?

R code

Python code

there is no difference between 0 and 0.0001?

[alexanderquispe]

2. If we change the number of iterations the results change?
3. When you regress the first case with LassoCV is because an ElasticNETCV with alpha =1 is a simple Lasso right ?

[alexanderquispe]

4. I don´t fully understand the meaning of this function.

I mean, we get g1 + m1 just from the two-equation above, but the while loop is running 5 times the same equation. But then, the return values of g1+m1 will be the results from iteration number 5 right ? what is the meaning of the loop then?
Do you have any idea?

[alexanderquispe]

5. If I understood well the second part Data Generating Process: Approximately Sparse + Small Dense Part, only differs with the first part in the beta composition?

Part 1

Part 2

[anzonyquispe]

1. The l1_ratio argument corresponds to sklearn package and alpha argument to glmnet r package. They have the same meaning in estimation. We have problems with estimations when l1_ratio is equal to zero. It is recommended a value close to zero.
2. I think the number of iterations does not change anything.
3. Yeah. That is right. They are the same. ElasticNETCV is the general model.
4. I have the same question. I do not have any idea.4.
5. Yeah, the author only changed beta.