GenWGP trains a generative flow to transport mass along Wasserstein gradient paths by optimizing a geometric action loss that encodes the full trajectory and equilibrium, matching reference solutions on Fokker-Planck and aggregation problems with roughly a dozen points.
For every π β₯ 0, there is a positive constant πΆπ such that sup(π‘,π₯)βΞ©Γ[0,π ] |β β π (π₯, π‘)| β€ πΆπ
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Generative Path-Finding Method for Wasserstein Gradient Flow
GenWGP trains a generative flow to transport mass along Wasserstein gradient paths by optimizing a geometric action loss that encodes the full trajectory and equilibrium, matching reference solutions on Fokker-Planck and aggregation problems with roughly a dozen points.