IEICE TRANSACTIONS on Information
Learning Curves in Learning with Noise
An Empirical Study
Summary :
In this paper, we apply the method of relating learning to hypothesis testing [6] to study average generalization performance of concept learning from noisy random training examples. A striking aspect of the method is that a learning problem with a so-called ill-disposed learning algorithm can equivalently be reduced to a simple one, and for this simple problem, even though a direct and exact calculation of the learning curves might still be impossible, a thorough empirical study can easily be performed. One of the main advantages of using the illdisposed algorithm is that it well models lower quality learning in real situations, and hence the result can provide useful implications as far as reliable generalization is concerned. We provide empirical formulas for the learning curves by simple functions of the noise rate and the sample size from a thorough empirical study, which smoothly incorporates the results from noise-free analysis and are quite accurate and adequate for practical applications when the noise rate is relatively small. The resulting learning curve bounds are directly related to the number of system weights and are not pessimistic in practice, and apply to learning settings not necessarily within the Bayesian framework.
- Publication
- IEICE TRANSACTIONS on Information Vol.E80-D No.1 pp.78-85
- Publication Date
- 1997/01/25
- Publicized
- Online ISSN
- DOI
- Type of Manuscript
- Category
- Bio-Cybernetics and Neurocomputing
Authors
Keyword
Latest Issue
Copyrights notice of machine-translated contents
The copyright of the original papers published on this site belongs to IEICE. Unauthorized use of the original or translated papers is prohibited. See IEICE Provisions on Copyright for details.
Cite this
Copy
Hanzhong GU, Haruhisa TAKAHASHI, "Learning Curves in Learning with NoiseAn Empirical Study" in IEICE TRANSACTIONS on Information,
vol. E80-D, no. 1, pp. 78-85, January 1997, doi: .
Abstract: In this paper, we apply the method of relating learning to hypothesis testing [6] to study average generalization performance of concept learning from noisy random training examples. A striking aspect of the method is that a learning problem with a so-called ill-disposed learning algorithm can equivalently be reduced to a simple one, and for this simple problem, even though a direct and exact calculation of the learning curves might still be impossible, a thorough empirical study can easily be performed. One of the main advantages of using the illdisposed algorithm is that it well models lower quality learning in real situations, and hence the result can provide useful implications as far as reliable generalization is concerned. We provide empirical formulas for the learning curves by simple functions of the noise rate and the sample size from a thorough empirical study, which smoothly incorporates the results from noise-free analysis and are quite accurate and adequate for practical applications when the noise rate is relatively small. The resulting learning curve bounds are directly related to the number of system weights and are not pessimistic in practice, and apply to learning settings not necessarily within the Bayesian framework.
URL: https://global.ieice.org/en_transactions/information/10.1587/e80-d_1_78/_p
Copy
@ARTICLE{e80-d_1_78,
author={Hanzhong GU, Haruhisa TAKAHASHI, },
journal={IEICE TRANSACTIONS on Information},
title={Learning Curves in Learning with NoiseAn Empirical Study},
year={1997},
volume={E80-D},
number={1},
pages={78-85},
abstract={In this paper, we apply the method of relating learning to hypothesis testing [6] to study average generalization performance of concept learning from noisy random training examples. A striking aspect of the method is that a learning problem with a so-called ill-disposed learning algorithm can equivalently be reduced to a simple one, and for this simple problem, even though a direct and exact calculation of the learning curves might still be impossible, a thorough empirical study can easily be performed. One of the main advantages of using the illdisposed algorithm is that it well models lower quality learning in real situations, and hence the result can provide useful implications as far as reliable generalization is concerned. We provide empirical formulas for the learning curves by simple functions of the noise rate and the sample size from a thorough empirical study, which smoothly incorporates the results from noise-free analysis and are quite accurate and adequate for practical applications when the noise rate is relatively small. The resulting learning curve bounds are directly related to the number of system weights and are not pessimistic in practice, and apply to learning settings not necessarily within the Bayesian framework.},
keywords={},
doi={},
ISSN={},
month={January},}
Copy
TY - JOUR
TI - Learning Curves in Learning with NoiseAn Empirical Study
T2 - IEICE TRANSACTIONS on Information
SP - 78
EP - 85
AU - Hanzhong GU
AU - Haruhisa TAKAHASHI
PY - 1997
DO -
JO - IEICE TRANSACTIONS on Information
SN -
VL - E80-D
IS - 1
JA - IEICE TRANSACTIONS on Information
Y1 - January 1997
AB - In this paper, we apply the method of relating learning to hypothesis testing [6] to study average generalization performance of concept learning from noisy random training examples. A striking aspect of the method is that a learning problem with a so-called ill-disposed learning algorithm can equivalently be reduced to a simple one, and for this simple problem, even though a direct and exact calculation of the learning curves might still be impossible, a thorough empirical study can easily be performed. One of the main advantages of using the illdisposed algorithm is that it well models lower quality learning in real situations, and hence the result can provide useful implications as far as reliable generalization is concerned. We provide empirical formulas for the learning curves by simple functions of the noise rate and the sample size from a thorough empirical study, which smoothly incorporates the results from noise-free analysis and are quite accurate and adequate for practical applications when the noise rate is relatively small. The resulting learning curve bounds are directly related to the number of system weights and are not pessimistic in practice, and apply to learning settings not necessarily within the Bayesian framework.
ER -
English
