Dropout Training is Distributionally Robust Optimal.
Jose H. BlanchetYang KangJosé Luis Montiel OleaViet Anh NguyenXuhui ZhangPublished in: J. Mach. Learn. Res. (2023)
Keyphrases
- early stopping
- robust optimization
- online learning
- dynamic programming
- test set
- neural network
- optimal design
- training phase
- robust estimation
- optimal control
- optimal solution
- training examples
- probabilistic model
- training set
- data sets
- linear programming
- worst case
- mathematical programming
- training algorithm
- multiscale
- training data
- video sequences