Convergence of Adam for Non-convex Objectives: Relaxed Hyperparameters and Non-ergodic Case.
Meixuan HeYuqing LiangJinlan LiuDongpo XuPublished in: CoRR (2023)
Keyphrases
- hyperparameters
- cross validation
- model selection
- bayesian framework
- support vector
- convex sets
- closed form
- em algorithm
- maximum a posteriori
- random sampling
- bayesian inference
- gaussian process
- convex optimization
- missing values
- regularization parameter
- gaussian processes
- prior information
- parameter settings
- incremental learning
- data sets
- sample size
- higher order
- high dimensional
- multiscale