Rewriting History with Inverse RL: Hindsight Inference for Policy Improvement.
Ben EysenbachXinyang GengSergey LevineRuss R. SalakhutdinovPublished in: NeurIPS (2020)
Keyphrases
- action selection
- optimal policy
- reinforcement learning
- markov decision process
- bayesian networks
- markov decision processes
- partially observable domains
- belief networks
- reinforcement learning problems
- significant improvement
- rl algorithms
- inference process
- control policy
- model free reinforcement learning
- state space
- policy search
- actor critic
- evidential reasoning
- action space
- learning algorithm
- model free
- bayesian inference
- function approximation
- integrity constraints