Self-Tuning Networks: Bilevel Optimization of Hyperparameters using Structured Best-Response Functions.
Matthew MacKayPaul VicolJonathan LorraineDavid DuvenaudRoger B. GrossePublished in: ICLR (Poster) (2019)
Keyphrases
- hyperparameters
- model selection
- cross validation
- grid search
- closed form
- random sampling
- bayesian inference
- maximum likelihood
- bayesian framework
- gaussian process
- em algorithm
- maximum a posteriori
- support vector
- prior information
- parameter optimization
- gaussian processes
- noise level
- sample size
- incremental learning
- incomplete data
- bilevel programming
- generalization ability
- lower bound
- image segmentation