Scalable Gradient-Based Tuning of Continuous Regularization Hyperparameters.
Jelena LuketinaTapani RaikoMathias BerglundKlaus GreffPublished in: ICML (2016)
Keyphrases
- hyperparameters
- regularization parameter
- prior information
- model selection
- cross validation
- bayesian inference
- closed form
- bayesian framework
- parameter settings
- random sampling
- support vector
- gaussian process
- gaussian processes
- posterior distribution
- noise level
- sample size
- maximum a posteriori
- em algorithm
- incremental learning
- maximum likelihood
- parameter optimization
- incomplete data
- missing values
- parameter space
- image restoration
- probabilistic model
- image reconstruction
- grid search
- computer vision
- high dimensional data
- support vector machine
- lower bound
- multiscale