papersSEP 11 04:00 UTC
Monotone Neural Policy Iteration Method for High-Dimensional HJB Equations
The paper studies a neural semi-discrete scheme for solving high-dimensional first-order Hamilton-Jacobi-Bellman equations, assuming either known or learned dynamics. Monotonicity is achieved by combining centered differences with an artificial viscosity term that scales linearly with the mesh size. The authors evaluate the resulting monotone operator within a policy iteration framework.
Hamilton-Jacobi-Bellman equationsartificial viscosityhigh-dimensional partial differential equationsmonotone schemesneural policy iterationpolicy iteration
COVERAGE · 2 REPORTS · LINKS GO TO THE ORIGINAL OUTLETS
arXiv cs.LGMonotone Neural Policy Iteration for High-Dimensional First-Order Hamilton--Jacobi--Bellman Equations ↗SEP 11 04:00 UTC
arXiv cs.AIMonotone Neural Policy Iteration for High-Dimensional First-Order Hamilton--Jacobi--Bellman Equations ↗SEP 12 04:00 UTC