Shaping Rewards for Reinforcement Learning with Imperfect Demonstrations using Generative Models.
Yuchen WuMelissa MozifianFlorian ShkurtiPublished in: CoRR (2020)
Keyphrases
- generative model
- reinforcement learning
- reward shaping
- reward function
- probabilistic model
- reinforcement learning algorithms
- mixture model
- complex domains
- discriminative models
- markov decision processes
- state space
- em algorithm
- discriminative learning
- semi supervised
- hidden variables
- hierarchical hidden markov models
- prior knowledge
- conditional random fields
- learning algorithm
- generative and discriminative models
- topic models
- learning process
- optimal policy
- expectation maximization
- representational power
- machine learning
- object categories
- supervised learning
- probability distribution
- hidden markov models
- deep belief networks
- naive bayes models
- training data