CAESAR: Enhancing Federated RL in Heterogeneous MDPs through Convergence-Aware Sampling with Screening.
Hei Yi MakFlint Xiaofeng FanLuca A. LanzendörferCheston TanWei Tsang OoiRoger WattenhoferPublished in: CoRR (2024)
Keyphrases
- reinforcement learning
- markov decision processes
- optimal policy
- stochastic shortest path
- state space
- state and action spaces
- distributed information systems
- markov decision process
- policy iteration
- function approximation
- action space
- reinforcement learning algorithms
- markov decision problems
- finite state
- dynamic programming
- multi agent
- convergence rate
- convergence speed
- random sampling
- continuous state and action spaces
- policy search
- policy evaluation
- digital libraries
- sample size
- markov games
- model free
- rl algorithms
- reward function
- least squares
- approximate policy iteration
- infinite horizon
- finite horizon
- monte carlo
- planning under uncertainty
- partially observable
- theoretical justification
- temporal difference
- linear programming
- machine learning