Towards Governing Agent's Efficacy: Action-Conditional β-VAE for Deep Transparent Reinforcement Learning.
John YangGyujeong LeeMinsung HyunSimyung ChangNojun KwakPublished in: CoRR (2018)
Keyphrases
- action selection
- reinforcement learning
- state action
- agent learns
- agent receives
- reward shaping
- multi agent
- action space
- state space
- action selection mechanism
- temporal difference
- partially observable
- practical reasoning
- partially observable domains
- learning agent
- reinforcement learning algorithms
- reward signal
- partial observations
- decision making
- function approximation
- autonomous agents
- reward function
- state abstraction
- markov decision processes
- machine learning
- mobile agents
- belief nets
- multiagent systems
- multi agent environments
- learning capabilities
- learning agents
- multi agent systems
- discounted reward
- model free
- intelligent agents
- plan execution
- policy iteration
- exploration strategy
- evaluation function
- function approximators
- internal state
- single agent
- human users
- communicative acts
- dynamic programming
- software agents