Deeper or Wider: A Perspective from Optimal Generalization Error with Sobolev Loss.
Yahong YangJuncai HePublished in: CoRR (2024)
Keyphrases
- feature set
- generalization error
- conditional expectation
- classification accuracy
- cross validation
- training set
- model selection
- bayes optimal
- upper bound
- active learning
- learning algorithm
- training data
- binary classification
- classification error
- training error
- worst case
- target function
- sample size
- linear classifiers
- sample complexity
- learning machines
- subspace information criterion
- perceptron learning
- supervised learning
- risk minimization
- decision trees
- data sets
- decision theory
- training process
- optimal solution
- low error
- learning rate