Generalization theory and generalization methods for neural networks

Zidonghua Xuebao/Acta Automatica Sinica

Volumn 27, Issue 6, 2001, Pages 806-815

  Wei, Hai Kun ^a   Xu, Si Xin ^a   Song, Wen Zhong ^a  

^a SOUTHEAST UNIVERSITY   (China)

Author keywords

Generalization ability; Generalization methods; Generalization theory; Neural networks

Indexed keywords

ALGORITHMS; CONVERGENCE OF NUMERICAL METHODS; MULTILAYER NEURAL NETWORKS; SAMPLING; TOPOLOGY;

GENERALIZATION ABILITY; GENERALIZATION METHOD; GENERALIZATION THEORY; SAMPLE COMPLEXITY; STRUCTURE COMPLEXITY;

FEEDFORWARD NEURAL NETWORKS;

EID: 0035502804     PISSN: 02544156     EISSN: None     Source Type: Journal    
DOI: None     Document Type: Article

References (67)

1. A theory of learnable
2. On the uniform convergence of relative frequencies of events to their probabilities
3. What size net gives valid generalization
4. Sample size requirements of feedforward neural networks classifiers
5. Generalization in feedforward networks
6. Sample size requirements for feedforward neural networks
7. Neural networks with quadratic VC-dimension
8. Neural nets with superlinear VC-dimension
9. Measuring the VC-dimension of a learning machine
10. MLP can provably generalize much better than VC-bound indicate
11. For valid generalization, the size of the weight is more important than the size of the network
12. The effective number of parameters: An analysis of generalization and regularization in nonlinear learning system
13. Approximation and estimation bounds for artificial neural networks
14. Neural networks and bias/variance dilemma
15. On the relationship between generalization error, hypothesis complexity, and sample complexity for radial basis function
16. Network generalization differences quantified
17. Improving generalization with active learning
18. Learning from hints in neural networks
19. A method of learning from Hints
20. Bayesian interpolation
21. Bayesian regularization and pruning using a laplace prior
22. Hints and the VC-dimension
23. Capacity control in classifiers for pattern recognition
24. Networks for approximation and learning
25. Generalization by weight-elimination with application to forcasting
26. Regularization theory and neural networks architecture
27. Back propagation is sensitive to initial conditions
28. Initialization, back-propagation and generalization of feed-forward classifiers
29. Learning and convergence analysis of neural-type structured networks
30. How initial conditions affect generalization performance in large networks
31. Temporal evolution of generalization during learning in linear networks
32. Generalization dynamics in LMS trained linear network
33. Optimal stopping and effective machine complexity in learning
34. Temporal dynamics of generalization in neural networks
35. Pruning algorithms: A survey
36. Constructive algorithms for structure learning in feedforward neural networks for regressing problems
37. Genetic evolution of the topology and weight distribution of neural networks
38. Network information criterion-determining the number of hidden units for an artificial neural network model
39. Neural networks algorithm that learn in polynomial time from examples and queries
40. Information-based objective function for active data selection
41. Creating artificial neural networks that generalization
42. Training with noise is equivalent to Tikhonov regularization
43. The effect of adding noise during backpropagation training on a generalization performance
44. Improving generalization through multiple output unit
45. Synergy of clustering multiple back propagation networks
46. Randomness in generalization ability: A source to improve it
47. Chaitin-kolmogorov complexity and generalization in neural networks
48. Engineering multivertion neural-net systems
49. Financial application of learning from hints
50. Including hints in training neural networks
51. Weight smoothing to improve network generalization
52. Overtraining, regularization and searching for a minimum, with application to neural networks
53. No free lunch for earlystopping
54. A numerical study onlearning curves in stochastic multilayer feedforward networks
55. Asymptotic statistical theory of overtraining and cross-validation
56. A bound on the error of cross validation using the approximation and estimation rates, with consequences for the training-test split
57. Decision theoretic generalization of the PAC model for neural net and other learning application
58. The lack of a priori distinctions between learning algorithms
59. The existence of a priori distinctions between learning algorithms