Machine Learning's Dropout Training is Distributionally Robust Optimal.
Jose H. BlanchetYang KangJose Luis Montiel OleaViet Anh NguyenXuhui ZhangPublished in: CoRR (2020)
Keyphrases
- machine learning
- robust optimization
- dynamic programming
- information extraction
- training set
- supervised learning
- training algorithm
- inductive learning
- knowledge acquisition
- learning algorithm
- learning machines
- partial occlusion
- closed form
- training examples
- computationally efficient
- text classification
- text mining
- active learning
- data analysis
- pattern recognition
- training samples
- computational intelligence
- test set
- expert systems
- training process
- reinforcement learning
- early stopping