On Assessing The Safety of Reinforcement Learning algorithms Using Formal Methods.
Paulina Stevia Nouwou MindomAmin NikanjamFoutse KhomhJohn MullinsPublished in: QRS (2021)
Keyphrases
- formal methods
- reinforcement learning algorithms
- safety critical
- reinforcement learning
- state space
- markov decision processes
- model free
- reinforcement learning problems
- model checking
- eligibility traces
- learning algorithm
- temporal difference
- knowledge based systems
- formal specification
- function approximation
- software engineering
- reinforcement learning methods
- artificial intelligence
- partially observable environments
- machine learning
- formal specification language
- supervised learning
- reward function
- development lifecycle
- dynamic systems
- markov models
- markov chain
- hidden markov models
- expert systems
- knowledge base
- reward shaping