Reward poisoning attacks in deep reinforcement learning based on exploration strategies.
Kanting CaiXiangbin ZhuZhaolong HuPublished in: Neurocomputing (2023)
Keyphrases
- reinforcement learning
- exploration strategy
- exploration exploitation
- state space
- function approximation
- action selection
- multi agent
- optimal policy
- learning agents
- learning algorithm
- average reward
- search strategies
- machine learning
- model free
- reinforcement learning algorithms
- partially observable environments
- malicious attacks
- active exploration
- reward function
- model based reinforcement learning
- reinforcement learning methods
- autonomous learning
- eligibility traces
- exploration exploitation tradeoff
- total reward
- malicious users
- function approximators
- countermeasures
- markov decision processes