Beyond Optimism: Exploration With Partially Observable Rewards.
Simone ParisiAlireza KazemipourMichael BowlingPublished in: CoRR (2024)
Keyphrases
- partially observable
- markov decision processes
- reinforcement learning
- reward function
- fully observable
- state space
- partial observability
- decision problems
- dynamical systems
- markov decision problems
- partial observations
- action selection
- finite state
- infinite horizon
- partially observable domains
- hidden state
- optimal policy
- bandit problems
- dynamic programming
- partially observable environments
- belief state
- learning algorithm
- orders of magnitude
- np hard
- belief space
- action models
- reinforcement learning algorithms
- policy iteration
- partially observable markov decision process
- average cost
- evolutionary algorithm
- optimal control
- machine learning