An approach to guaranteeing generalisation in neural networks

Neural Networks

Volumn 14, Issue 8, 2001, Pages 1035-1048

  Polhill, J Gary ^a   Weir, Michael K ^b  

^a SCOTTISH CROP RESEARCH INSTITUTE   (United Kingdom)

^b UNIVERSITY OF ST ANDREWS   (United Kingdom)

Author keywords

Binary outputs; Exhaustive search; Feed forward neural networks; Global minimum error; Guaranteed generalisation; Target reversal

Indexed keywords

CONVERGENCE OF NUMERICAL METHODS; ERROR ANALYSIS;

GLOBAL MINIMUM ERROR ANALYSIS;

NEURAL NETWORKS;

ANALYTICAL ERROR; ARTICLE; ARTIFICIAL INTELLIGENCE; ARTIFICIAL NEURAL NETWORK; INFORMATION PROCESSING; MATHEMATICAL ANALYSIS; PRIORITY JOURNAL; SIGNAL PROCESSING; TECHNIQUE; VALIDATION PROCESS;

EID: 0034842475     PISSN: 08936080     EISSN: None     Source Type: Journal    
DOI: 10.1016/S0893-6080(01)00061-2     Document Type: Article

References (41)

1. The Vapnik-Chervonenkis dimension: Information versus complexity in learning
2. What size net gives valid generalization?
3. No free lunch for early stopping
4. Neural networks: A review from a statistical perspective
5. Large automatic learning, rule extraction, and generalization
6. On the approximate realization of continuous mappings by neural networks
7. Neural networks and the bias/variance dilemma
8. What is Occam's razor?
9. Learning generalization by validation set
10. Bounds on the sample complexity of Bayesian learning using information theory and the VC dimension
11. Multilayer feedforward networks are universal approximators
12. Neural network pruning by fusing hidden layer units
13. A bound on the error of cross validation using the approximation and estimation rates, with consequences for the training-test split
14. Algorithmic stability and sanity-check bounds for leave-one-out cross-validation
15. A new method to increase the margin of multilayer perceptrons
16. An introduction to computing with neural nets
17. Bayesian interpolation
18. A practical Bayesian framework for backpropagation networks
19. The evidence framework applied to classification networks
20. The appeal of parallel distributed processing
21. Estimating MLP generalisation ability without a test set using fast, approximate leave-one-out cross-validation
22. Simplifying neural networks by soft weight-sharing
23. Verification, validation, and confirmation of numerical models in the earth sciences
24. Generalization performance of Bayes optimal classification algorithm for learning a perceptron
25. Automatic early stopping using cross validation: Quantifying the criteria
26. Two strategies to avoid overfitting in feedforward neural networks
27. Discovering neural nets with low Kolmogorov complexity and high generalization capability
28. Exhaustive learning
29. Creating artificial neural networks that generalize
30. Neural network constructive algorithms: Trading generalization for learning efficiency?
31. The geometry of multi-layer perceptron solutions
32. Uniqueness of the weights for minimal feedforward nets with a given input-output map
33. The myth of Occam's razor
34. A theory of cross-validation error
35. On the uniform convergence of relative frequencies of events to their probabilities
36. Back-propagation, weight-elimination, and time series prediction
37. Bidirectional convergence: A cognitive approach to generalisation
38. Neural bidirectional convergence: A method for concept learning in neural networks and symbolic AI