Constrained Reinforcement Learning Has Zero Duality Gap.

Santiago Paternain Luiz F. O. Chamon Miguel Calvo-Fullana Alejandro Ribeiro

Published in: NeurIPS (2019)

Keyphrases

duality gap
reinforcement learning
primal dual
augmented lagrangian
set covering problem
inequality constraints
convex programming
lagrange multipliers
cost function
saddle point
optimal solution
convex optimization
integer programming
optimality conditions