REVEAL-IT: REinforcement learning with Visibility of Evolving Agent poLicy for InTerpretability.
Shuang AoSimon KhanHaris AzizFlora D. SalimPublished in: CoRR (2024)
Keyphrases
- reinforcement learning
- action selection
- reward function
- markov decision process
- partially observable
- agent learns
- state action
- optimal policy
- agent receives
- multi agent
- policy search
- markov decision processes
- state space
- action space
- markov decision problems
- multiagent systems
- state and action spaces
- partially observable markov decision process
- reward shaping
- total reward
- actor critic
- partially observable environments
- policy iteration
- learning agent
- function approximation
- state abstraction
- partially observable domains
- selective perception
- decision theoretic
- approximate dynamic programming
- reward signal
- optimal control
- learning agents
- multiple agents
- learning capabilities
- decision making
- learning algorithm
- control policies
- agent architecture
- exploration strategy
- policy gradient
- intelligent agents
- reinforcement learning algorithms
- decision problems
- continuous state
- prediction accuracy
- control policy
- infinite horizon
- model free
- temporal difference
- partially observable markov decision processes
- machine learning
- multi agent systems
- discounted reward
- mobile agents
- dynamical systems
- reinforcement learning agents
- multiagent reinforcement learning
- reinforcement learning problems
- multi agent environments
- inverse reinforcement learning
- autonomous agents
- software agents
- continuous state spaces
- autonomous learning
- rl algorithms
- function approximators