Switching Isotropic and Directional Exploration with Parameter Space Noise in Deep Reinforcement Learning.
Izumi KarinoKazutoshi TanakaRyuma NiiyamaYasuo KuniyoshiPublished in: CoRR (2018)
Keyphrases
- parameter space
- reinforcement learning
- noise model
- starting points
- active exploration
- image space
- exploration strategy
- parameter values
- signal to noise ratio
- multivariate data
- high dimensional
- action selection
- noisy data
- autonomous learning
- model based reinforcement learning
- noise reduction
- search space
- gaussian noise
- directional information
- exploration exploitation
- reinforcement learning algorithms
- machine learning
- function approximation
- model free
- noise level
- markov decision processes
- linear transformations
- hough space
- singularity detection
- likelihood function
- transfer learning
- basis functions
- optimal policy
- state space
- exploration exploitation tradeoff