Differentially Private Reward Functions for Multi-Agent Markov Decision Processes.
Alexander BenvenutiCalvin HawkinsBrandon FallinBo ChenBrendan J. BialyMiriam DennisMatthew T. HalePublished in: CoRR (2023)
Keyphrases
- differentially private
- reward function
- markov decision processes
- multi agent
- reinforcement learning
- planning under uncertainty
- differential privacy
- state space
- multiple agents
- partially observable
- reinforcement learning algorithms
- finite state
- optimal policy
- dynamic programming
- policy iteration
- factored mdps
- markov decision process
- average reward
- transition matrices
- multi agent systems
- decision processes
- average cost
- hierarchical reinforcement learning
- stochastic games
- privacy preserving
- infinite horizon
- function approximation
- action space
- partially observable markov decision processes
- state abstraction
- objective function
- random walk
- learning algorithm
- model free