An Abstraction-based Method to Verify Multi-Agent Deep Reinforcement-Learning Behaviours.
Pierre El MqirmiFrancesco BelardinelliBorja G. LeónPublished in: CoRR (2021)
Keyphrases
- multi agent
- reinforcement learning
- high accuracy
- high precision
- optimization method
- segmentation method
- prior knowledge
- synthetic data
- cost function
- dynamic programming
- experimental evaluation
- detection method
- computational cost
- detection algorithm
- optimization algorithm
- unsupervised learning
- classification accuracy
- genetic algorithm
- significant improvement
- objective function
- training data
- feature selection