Reward Certification for Policy Smoothed Reinforcement Learning.
Ronghui MuLeandro Soriano MarcolinoTianle ZhangYanghao ZhangXiaowei HuangWenjie RuanPublished in: CoRR (2023)
Keyphrases
- reinforcement learning
- optimal policy
- reward function
- partially observable environments
- total reward
- policy search
- policy gradient
- average reward
- action selection
- partially observable
- reinforcement learning algorithms
- eligibility traces
- actor critic
- markov decision problems
- markov decision process
- agent learns
- control policy
- markov decision processes
- state action
- state space
- function approximators
- inverse reinforcement learning
- policy iteration
- function approximation
- reinforcement learning problems
- control policies
- approximate dynamic programming
- policy evaluation
- action space
- partially observable markov decision processes
- temporal difference
- model free
- rl algorithms
- optimal control
- long run
- decision problems
- reinforcement learning methods
- learning algorithm
- continuous state
- expected reward
- multi agent
- infinite horizon
- state and action spaces
- discounted reward
- partially observable domains
- continuous state spaces
- learning process
- agent receives
- finite state
- multiple agents
- control problems