Regret Minimization for Partially Observable Deep Reinforcement Learning.
Peter H. JinSergey LevineKurt KeutzerPublished in: ICLR (Workshop) (2018)
Keyphrases
- partially observable
- reinforcement learning
- regret minimization
- decision problems
- game theoretic
- state space
- markov decision processes
- partial observability
- nash equilibrium
- partially observable environments
- partially observable domains
- action models
- function approximation
- markov decision problems
- partial observations
- dynamical systems
- reward function
- belief state
- reinforcement learning algorithms
- learning algorithm
- hidden state
- optimal policy
- markov decision process
- temporal difference
- multi agent learning
- fully observable
- multi agent
- model free
- dynamic programming
- infinite horizon
- learning capabilities
- heuristic search
- search space
- transfer learning