L0 Trend Filtering with Automatic Knot Detection

Professor Li, Guodong   (Principal Investigator (PI))

30

2019-09-01

332261

L0 Trend Filtering with Automatic Knot Detection

Active set search, Automatic knot detection, Primal-dual formulation, Superset consistency, Trend filtering

Probability & Statistics

Physical Sciences (P)

17306519

General Research Fund (GRF)

2019

Completed

1 to further develop the active set search algorithm with improved performance for automatic knot detection; 2 to establish the connection of L0 trend filtering with adaptive regression splines with homogeneous and inhomogeneous smoothness; 3 to derive the conditions for the improved algorithm with provable guarantee of finite-step convergence and local consistency; 4 to develop real data applications of L0 trend filtering for change point detection and time series segmentation; 5 to develop a composite L0 trend filtering method with potential application to DNA copy number variation detection in genome analysis.