Material covered

  • Balancing data fit and model complexity with Ridge regression
  • Alternative way to write optimization as finding parameters within disk
  • Behavior of model weights as disk radius is changed (demo)
  • "Evidence-sharing" and geometric reasons for non-sparse solutions
  • Lasso regression: replacing L2 model penalty with L1 model penalty
  • Casting Lasso regression as a scilab/matlab quadratic program
  • Behavior of model weights as L1 limit is changed (demo)
  • Reasons for Lasso giving more sparse solutions
  • Implicit feature selection effect
  • Handling non-linear regression by lifting data points to higher dimension
  • Classification with small discrete label sets
  • Why classification loss is fundamentally different from regression

Last modified: Saturday, 25 August 2007, 10:15 AM