RMSprop converges with proper hyper-parameter.
Naichen ShiDawei LiMingyi HongRuoyu SunPublished in: ICLR (2021)
Keyphrases
- hyperparameters
- model selection
- cross validation
- random sampling
- bayesian inference
- bayesian framework
- support vector
- closed form
- em algorithm
- maximum a posteriori
- parameter settings
- prior information
- sample size
- maximum likelihood
- incomplete data
- noise level
- image processing
- feature selection
- generative model
- multiresolution
- evolutionary algorithm
- optimal solution
- bayesian networks