Deep Transformer Q-Networks for Partially Observable Reinforcement Learning.
Kevin EsslingerRobert PlattChristopher AmatoPublished in: CoRR (2022)
Keyphrases
- partially observable
- reinforcement learning
- state space
- partial observability
- markov decision processes
- dynamical systems
- partially observable environments
- partially observable domains
- decision problems
- infinite horizon
- hidden state
- action models
- markov decision problems
- social networks
- partial observations
- function approximation
- machine learning
- reinforcement learning algorithms
- belief space
- learning algorithm
- reward function
- model free
- partially observable markov decision processes
- state variables
- partially observable markov decision process
- fully observable
- belief state
- planning domains
- finite state
- optimal policy
- dynamic programming
- optimal control
- transfer learning
- orders of magnitude
- decision makers
- hidden markov models