Action Noise in Off-Policy Deep Reinforcement Learning: Impact on Exploration and Performance.
Jakob J. HollensteinSayantan AuddyMatteo SaverianoErwan RenaudoJustus H. PiaterPublished in: CoRR (2022)
Keyphrases
- reinforcement learning
- action selection
- active exploration
- action space
- machine learning
- state action
- markov decision processes
- function approximation
- model based reinforcement learning
- partially observable domains
- reward shaping
- random noise
- noise reduction
- noisy data
- learning algorithm
- exploration exploitation
- autonomous learning
- temporal difference
- deep learning
- signal to noise ratio
- noise level
- multi agent
- image noise
- reinforcement learning algorithms
- noise free
- missing data
- optimal policy
- model free
- exploration strategy
- input data
- gaussian noise
- state space