Curvature from Potts Fisher information on labeled k-NN graphs yields training subsets that beat random and entropy sampling on 60+ datasets under a k-NN classifier.
Xiaojin Zhu, Zoubin Ghahramani, and John Lafferty
1 Pith paper cite this work, alongside 696 external citations. Polarity classification is still indexing.
1
Pith paper citing it
696
external citations · OpenAlex
fields
cs.LG 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
CuBAS: Information Geometric Curvature-Based Adaptive Sampling for Supervised Classification
Curvature from Potts Fisher information on labeled k-NN graphs yields training subsets that beat random and entropy sampling on 60+ datasets under a k-NN classifier.