Shaping Rewards for Reinforcement Learning with Imperfect Demonstrations using Generative Models.
Yuchen WuMelissa MozifianFlorian ShkurtiPublished in: ICRA (2021)
Keyphrases
- generative model
- reinforcement learning
- reward shaping
- reward function
- probabilistic model
- reinforcement learning algorithms
- discriminative learning
- mixture model
- state space
- complex domains
- discriminative models
- em algorithm
- maximum entropy principle
- markov decision processes
- semi supervised
- conditional random fields
- machine learning
- prior knowledge
- generative and discriminative models
- object categories
- transfer learning
- optimal policy
- expectation maximization
- hierarchical hidden markov models
- representational power
- learning process
- hidden variables
- image processing