ID W8: Nearest-Neighbor-Based Active Learning for Rare Category Detection
Jingrui He, Jaime Carbonell How to Find NIPS?
Bad News

· No Labeled Examples · NIPS are Rare Classes · Class Priors are Known · Majority Class is Smooth · Rare Classes are Clustered

Good News

6% of the labeled points needed by random sampling
200 190 180 170 160 150 140 130 120 120 140 160 180 200 220 200 190 180 170 160 150 140 130 120 120 140 160 180 200 220

Local Density

Difference in Density

NIPS Found!