Double Duality: Variational Primal-Dual Policy Optimization for Constrained Reinforcement Learning.
Zihao LiBoyi LiuZhuoran YangZhaoran WangMengdi WangPublished in: CoRR (2024)
Keyphrases
- primal dual
- saddle point
- linear programming
- reinforcement learning
- duality gap
- optimal policy
- linear program
- convex programming
- affine scaling
- dual formulation
- convex optimization
- semidefinite programming
- interior point methods
- variational inequalities
- line search
- algorithm for linear programming
- linear programming problems
- policy search
- convex optimization problems
- simplex algorithm
- interior point
- interior point algorithm
- nonlinear programming
- approximation algorithms
- markov decision process
- convergence rate
- quadratic programming
- dynamic programming
- function approximators
- image segmentation
- lagrange multipliers
- optimization problems
- reward function
- semidefinite
- temporal difference
- simplex method
- action selection
- markov decision processes
- machine learning
- function approximation
- state space
- feasible solution
- column generation
- special case
- constrained optimization
- motion estimation
- policy gradient
- maximum margin
- policy iteration