In order to obtain better generalization performance in supervised learning, model parameters should be determined appropriately, i.e., they should be determined so that the generalization error is minimized. However, since the generalization error is inaccessible in practice, the model parameters are usually determined so that an estimator of the generalization error is minimized. The regularized subspace information criterion (RSIC) is such a generalization error estimator for model selection. RSIC includes an additional regularization parameter and it should be determined appropriately for better model selection. A meta-criterion for determining the regularization parameter has also been proposed and shown to be useful in practice. In this paper, we show that there are several drawbacks in the existing meta-criterion and give an alternative meta-criterion that can solve the problems. Through simulations, we show that the use of the new meta-criterion further improves the model selection performance.
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.
Copy
Yasushi HIDAKA, Masashi SUGIYAMA, "A New Meta-Criterion for Regularized Subspace Information Criterion" in IEICE TRANSACTIONS on Information,
vol. E90-D, no. 11, pp. 1779-1786, November 2007, doi: 10.1093/ietisy/e90-d.11.1779.
Abstract: In order to obtain better generalization performance in supervised learning, model parameters should be determined appropriately, i.e., they should be determined so that the generalization error is minimized. However, since the generalization error is inaccessible in practice, the model parameters are usually determined so that an estimator of the generalization error is minimized. The regularized subspace information criterion (RSIC) is such a generalization error estimator for model selection. RSIC includes an additional regularization parameter and it should be determined appropriately for better model selection. A meta-criterion for determining the regularization parameter has also been proposed and shown to be useful in practice. In this paper, we show that there are several drawbacks in the existing meta-criterion and give an alternative meta-criterion that can solve the problems. Through simulations, we show that the use of the new meta-criterion further improves the model selection performance.
URL: https://global.ieice.org/en_transactions/information/10.1093/ietisy/e90-d.11.1779/_p
Copy
@ARTICLE{e90-d_11_1779,
author={Yasushi HIDAKA, Masashi SUGIYAMA, },
journal={IEICE TRANSACTIONS on Information},
title={A New Meta-Criterion for Regularized Subspace Information Criterion},
year={2007},
volume={E90-D},
number={11},
pages={1779-1786},
abstract={In order to obtain better generalization performance in supervised learning, model parameters should be determined appropriately, i.e., they should be determined so that the generalization error is minimized. However, since the generalization error is inaccessible in practice, the model parameters are usually determined so that an estimator of the generalization error is minimized. The regularized subspace information criterion (RSIC) is such a generalization error estimator for model selection. RSIC includes an additional regularization parameter and it should be determined appropriately for better model selection. A meta-criterion for determining the regularization parameter has also been proposed and shown to be useful in practice. In this paper, we show that there are several drawbacks in the existing meta-criterion and give an alternative meta-criterion that can solve the problems. Through simulations, we show that the use of the new meta-criterion further improves the model selection performance.},
keywords={},
doi={10.1093/ietisy/e90-d.11.1779},
ISSN={1745-1361},
month={November},}
Copy
TY - JOUR
TI - A New Meta-Criterion for Regularized Subspace Information Criterion
T2 - IEICE TRANSACTIONS on Information
SP - 1779
EP - 1786
AU - Yasushi HIDAKA
AU - Masashi SUGIYAMA
PY - 2007
DO - 10.1093/ietisy/e90-d.11.1779
JO - IEICE TRANSACTIONS on Information
SN - 1745-1361
VL - E90-D
IS - 11
JA - IEICE TRANSACTIONS on Information
Y1 - November 2007
AB - In order to obtain better generalization performance in supervised learning, model parameters should be determined appropriately, i.e., they should be determined so that the generalization error is minimized. However, since the generalization error is inaccessible in practice, the model parameters are usually determined so that an estimator of the generalization error is minimized. The regularized subspace information criterion (RSIC) is such a generalization error estimator for model selection. RSIC includes an additional regularization parameter and it should be determined appropriately for better model selection. A meta-criterion for determining the regularization parameter has also been proposed and shown to be useful in practice. In this paper, we show that there are several drawbacks in the existing meta-criterion and give an alternative meta-criterion that can solve the problems. Through simulations, we show that the use of the new meta-criterion further improves the model selection performance.
ER -