Optimality of Thompson Sampling for Gaussian Bandits Depends on Priors.
Junya HondaAkimichi TakemuraPublished in: AISTATS (2014)
Keyphrases
- heavy tailed
- multi armed bandit
- random sampling
- sampling strategy
- optimal solution
- gaussian mixture model
- stochastic systems
- learned from training data
- maximum likelihood
- sample size
- prior information
- gaussian distribution
- sampling methods
- multi armed bandits
- e learning
- image prior
- bayesian framework
- parameter space
- monte carlo
- probability distribution
- prior knowledge