Critical Hyper-Parameters: No Random, No Cry.
Olivier BousquetSylvain GellyKarol KurachOlivier TeytaudDamien VincentPublished in: CoRR (2017)
Keyphrases
- hyperparameters
- model selection
- cross validation
- closed form
- random sampling
- bayesian framework
- bayesian inference
- support vector
- em algorithm
- gaussian process
- prior information
- bayesian methods
- maximum likelihood
- posterior distribution
- noise level
- sample size
- incremental learning
- variational bayes
- missing values
- maximum a posteriori
- parameter settings
- incomplete data
- conjugate priors
- parameter space
- markov random field
- dimensionality reduction