Increasing the Action Gap: New Operators for Reinforcement Learning.
Marc G. BellemareGeorg OstrovskiArthur GuezPhilip S. ThomasRémi MunosPublished in: CoRR (2015)
Keyphrases
- reinforcement learning
- action selection
- partially observable domains
- action space
- model free
- function approximation
- reward shaping
- reinforcement learning algorithms
- state action
- markov decision processes
- optimal policy
- robotic control
- multi agent
- temporal difference
- learning algorithm
- learning capabilities
- neural network
- transition model
- fitted q iteration
- continuous state
- function approximators
- human actions
- mobile robot
- knowledge base
- machine learning