Skip to content

Latest commit

 

History

History
8 lines (7 loc) · 462 Bytes

regression.md

File metadata and controls

8 lines (7 loc) · 462 Bytes

Influence and Leverage for active learning on regression problems.

Based on the statistical concepts of leverage of linear regression, we should be able to recommend points that are most useful to get a label.

  • What would be the GP version?
  • How it this related to D-optimal design?
  • Leverage additionally takes the residual error into account, and therefore needs access to the labels to compute. However a semi-supervised approach might be feasible.