CLEANing the reward: counterfactual actions to remove exploratory action noise in multiagent learning (extended abstract).
Chris HolmesParkerMatthew E. TaylorAdrian K. AgoginoKagan TumerPublished in: AAMAS (2014)
Keyphrases
- extended abstract
- multiagent learning
- reinforcement learning
- human actions
- action selection
- state action
- multi agent
- reasoning about actions
- agent behavior
- reward function
- multiagent systems
- internal state
- learning agents
- learning agent
- initial state
- resource allocation
- action recognition
- agent learns
- action space
- game theoretic
- reward signal
- multiple agents
- single agent
- multi agent learning
- state transitions
- agent receives
- function approximation
- long run
- mobile robot
- game theory
- human activities
- situation calculus