SAFE-RL: Saliency-Aware Counterfactual Explainer for Deep Reinforcement Learning Policies.
Amir SamadiKonstantinos KoufosKurt DebattistaMehrdad DianatiPublished in: CoRR (2024)
Keyphrases
- reinforcement learning
- optimal policy
- policy search
- control policies
- markov decision process
- partially observable markov decision processes
- reinforcement learning agents
- state space
- hierarchical reinforcement learning
- reward function
- total reward
- semi markov decision process
- markov decision processes
- reinforcement learning algorithms
- function approximation
- control policy
- markov decision problems
- fitted q iteration
- learning algorithm
- decision problems
- multi agent
- continuous state
- saliency map
- model free
- rl algorithms
- visual attention
- saliency detection
- finite state
- temporal difference
- average reward
- control problems
- long run
- dynamic programming
- approximate policy iteration
- action space
- machine learning
- reinforcement learning problems
- transfer learning
- partially observable
- complex domains
- learning problems
- learning agents
- visual saliency
- state abstraction
- autonomous learning
- multiagent reinforcement learning
- input image
- direct policy search
- policy iteration