Optimizing Millions of Hyperparameters by Implicit Differentiation.
Jonathan LorrainePaul VicolDavid DuvenaudPublished in: CoRR (2019)
Keyphrases
- hyperparameters
- model selection
- cross validation
- support vector
- bayesian inference
- closed form
- random sampling
- bayesian framework
- noise level
- gaussian process
- prior information
- maximum a posteriori
- em algorithm
- sample size
- maximum likelihood
- regularization parameter
- incremental learning
- incomplete data
- gaussian processes
- prior knowledge
- parameter space
- missing values
- parameter settings
- grid search
- np hard
- nearest neighbor
- active learning
- noisy images
- data streams
- markov random field
- machine learning