Offline Primal-Dual Reinforcement Learning for Linear MDPs.
Germano GabbianelliGergely NeuMatteo PapiniNneka OkoloPublished in: AISTATS (2024)
Keyphrases
- primal dual
- reinforcement learning
- markov decision processes
- linear programming
- affine scaling
- interior point methods
- convex optimization
- state space
- linear program
- convergence rate
- approximation algorithms
- simplex algorithm
- variational inequalities
- interior point algorithm
- linear programming problems
- function approximation
- algorithm for linear programming
- optimal policy
- function approximators
- semidefinite programming
- markov decision problems
- policy search
- partially observable
- continuous state and action spaces
- quadratic programming
- reinforcement learning algorithms
- duality gap
- markov decision process
- simplex method
- policy iteration
- reward function
- semidefinite
- action selection
- model free
- learning algorithm
- state and action spaces
- saddle point
- dynamic programming
- machine learning
- image processing
- dual formulation
- action space
- policy evaluation
- continuous state
- special case
- wavelet transform
- finite number