Rethinking the Hyperparameters for Fine-tuning.
Hao LiPratik ChaudhariHao YangMichael LamAvinash RavichandranRahul BhotikaStefano SoattoPublished in: CoRR (2020)
Keyphrases
- fine tuning
- hyperparameters
- model selection
- cross validation
- bayesian inference
- closed form
- random sampling
- support vector
- bayesian framework
- prior information
- noise level
- em algorithm
- gaussian process
- maximum likelihood
- sample size
- incremental learning
- gaussian processes
- regularization parameter
- maximum a posteriori
- missing values
- incomplete data
- fine tuned
- grid search
- parameter space
- noisy images
- error rate
- expectation maximization
- probabilistic model
- multiscale
- feature selection
- pairwise
- machine learning