Gradient-based learning algorithms with constant-error estimators: stability and convergence.
Arunselvan RamaswamyShalabh BhatnagarPublished in: CoRR (2016)
Keyphrases
- learning algorithm
- variance reduction
- low variance
- rates of convergence
- generalization error
- error rate
- iterative algorithms
- machine learning
- machine learning algorithms
- active learning
- algorithmic stability
- learning rate
- convergence rate
- membership and equivalence queries
- unbiased estimator
- stability analysis
- general loss functions
- learning problems
- supervised learning
- upper bound
- reinforcement learning
- training data
- convergence speed
- confidence intervals
- error analysis
- model free
- monte carlo
- training samples
- decision tree learning
- feature space
- genetic algorithm
- data sets