Deep Active Inference for Partially Observable MDPs.
Otto van der HimstPablo LanillosPublished in: CoRR (2020)
Keyphrases
- partially observable
- markov decision processes
- markov decision problems
- reinforcement learning
- state space
- decision problems
- dynamical systems
- partial observability
- infinite horizon
- partial observations
- reward function
- partially observable environments
- probabilistic planning
- action models
- finite state
- decision theoretic
- belief state
- policy iteration
- probabilistic inference
- dec pomdps
- markov decision process
- bayesian networks
- optimal policy
- reinforcement learning algorithms
- planning domains
- average cost
- probabilistic model
- average reward
- search space
- computational complexity
- dynamic bayesian networks