Notes on Lasso.jl
For CBIOMES Julia meeting discussion lead by @jtsiddons in October.
We look at using Lasso.jl
and GLMNet.jl
to perform penalised regression in Julia. I have created a quick dataset from the Narragansett Bay NABATS.org, from which we will use environmental and meteorlogical variables to estimate surface salinity in the bay. I will start with a simple linear regression using GLM.jl
.
- What is Lasso Regression? 1st order penalty on the regression coefficients.
- How do coefficients change with
$\lambda$ (plot)
2nd order penalty to the coefficients.
- How does changing
$\lambda$ and/or$\alpha$ affect the result
Use cross-validation using K-fold from MLBase
.