Information-Theoretic Regret Bounds for Gaussian Process Optimization in the Bandit Setting.
Niranjan SrinivasAndreas KrauseSham M. KakadeMatthias W. SeegerPublished in: IEEE Trans. Inf. Theory (2012)
Keyphrases
- information theoretic
- regret bounds
- gaussian process
- bregman divergences
- multi armed bandit
- mutual information
- lower bound
- online learning
- regression model
- bayesian framework
- semi supervised
- linear regression
- upper bound
- approximate inference
- model selection
- hyperparameters
- latent variables
- log likelihood
- kullback leibler divergence
- kl divergence
- closed form
- maximum likelihood