Regularisation: Ridge, Lasso, Elastic Net (and L1/L2 for logistic)

Fit Ridge, Lasso and ElasticNet with a range of alpha; plot coefficient paths; use penalty='l1'/'l2' in logistic regression. Concepts: L2 shrinkage; L1 sparsity (feature selection); Elastic Net mix Project A: L1 logistic regression: which features survive at C = 0.05? Compare with the EDA consensus shortlist. Project B: Ridge on many correlated lag/weather features; Lasso path shows which lags matter.