The paper defines a Riemannian geometry on policies via Wasserstein space and stationary distributions, then constructs a gradient flow for RL optimization using Otto's calculus with explicit gradient and Hessian computations.
Springer, 3rd edition
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
-
Wasserstein Formulation of Reinforcement Learning. An Optimal Transport Perspective on Policy Optimization
The paper defines a Riemannian geometry on policies via Wasserstein space and stationary distributions, then constructs a gradient flow for RL optimization using Otto's calculus with explicit gradient and Hessian computations.