A universal theorem on learning curves

Neural Networks

Volumn 6, Issue 2, 1993, Pages 161-166

  Amari, Shun ichi ^a  

^a UNIVERSITY OF TOKYO   (Japan)

Author keywords

Entropic error; Generalization error; Information gain; Learning curve; Universal theorem

Indexed keywords

AUTOMATA THEORY; ERROR ANALYSIS; MATHEMATICAL MODELS; NEURAL NETWORKS; STATISTICAL MECHANICS; THEOREM PROVING;

ENTROPIC ERROR; GENERAL NOISELESS DICHOTOMY MACHINES; GENERALIZATION ERROR; INFORMATION GAIN; LEARNING CURVES; UNIVERSAL THEOREM;

LEARNING SYSTEMS;

EID: 0027257001     PISSN: 08936080     EISSN: None     Source Type: Journal    
DOI: 10.1016/0893-6080(93)90013-M     Document Type: Article

References (15)

1. Theory of adaptive pattern classifiers
2. Four types of learning curves
3. Statistical theory of learning curves under entropic loss criterion
4. What size net gives valid generalization
5. Statistical theory of learning a rule
6. Bounds on the sample complexity of Bayesian learning using information theory and the VC dimension
7. Predicting {0, 1} functions on randomly drawn points
8. Learning processes in neural networks
9. A statistical approach to learning and generalization in layered neural networks
10. Network information criterion—Determining the number of hidden units for an artificial neural network model
11. Calculation of the learning curve of Bayes optimal classification algorithm for learning perceptron with noise
12. Stochastic complexity and modeling
13. Learning from examples in large neural networks
14. A theory of the learnable
15. A loss bound model for on-line stochastic prediction strategies