s-OTDD is a near-linear-time dataset distance that projects labels via scaled moments of their feature distributions and matches OTDD's correlations at a fraction of the cost.
Hyperbolic sliced- W asserstein via geodesic and horospherical projections
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Lightspeed Geometric Dataset Distance via Sliced Optimal Transport
s-OTDD is a near-linear-time dataset distance that projects labels via scaled moments of their feature distributions and matches OTDD's correlations at a fraction of the cost.