Sparse Regression and Trees
Sparse models and regression trees reduce complexity in different ways: one limits the active coefficients, while the other partitions the predictor space.
Read the overview before LASSO and coordinate descent. Follow solution paths to understand selection, then study regression trees and the interpretation of fitted results.
Sparse Regression and Trees
Overview and model setup.
LASSO
Use an absolute-value penalty for sparse coefficients.
Coordinate Descent
Update one coefficient at a time.
Solution Paths and Selection
Follow coefficients as the penalty changes.
Regression Trees
Partition predictors into regions and prune the tree.
Reading Model Results
Interpret coefficients, validation curves, and tree summaries.