The Hilbert-Schmidt independence criterion (HSIC) is a kernel-based statistical independence measure that can be computed very efficiently. However, it requires us to determine the kernel parameters heuristically because no objective model selection method is available. Least-squares mutual information (LSMI) is another statistical independence measure that is based on direct density-ratio estimation. Although LSMI is computationally more expensive than HSIC, LSMI is equipped with cross-validation, and thus the kernel parameter can be determined objectively. In this paper, we show that HSIC can actually be regarded as an approximation to LSMI, which allows us to utilize cross-validation of LSMI for determining kernel parameters in HSIC. Consequently, both computational efficiency and cross-validation can be achieved.
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
Masashi SUGIYAMA, Makoto YAMADA, "On Kernel Parameter Selection in Hilbert-Schmidt Independence Criterion" in IEICE TRANSACTIONS on Information,
vol. E95-D, no. 10, pp. 2564-2567, October 2012, doi: 10.1587/transinf.E95.D.2564.
Abstract: The Hilbert-Schmidt independence criterion (HSIC) is a kernel-based statistical independence measure that can be computed very efficiently. However, it requires us to determine the kernel parameters heuristically because no objective model selection method is available. Least-squares mutual information (LSMI) is another statistical independence measure that is based on direct density-ratio estimation. Although LSMI is computationally more expensive than HSIC, LSMI is equipped with cross-validation, and thus the kernel parameter can be determined objectively. In this paper, we show that HSIC can actually be regarded as an approximation to LSMI, which allows us to utilize cross-validation of LSMI for determining kernel parameters in HSIC. Consequently, both computational efficiency and cross-validation can be achieved.
URL: https://global.ieice.org/en_transactions/information/10.1587/transinf.E95.D.2564/_p
Copy
@ARTICLE{e95-d_10_2564,
author={Masashi SUGIYAMA, Makoto YAMADA, },
journal={IEICE TRANSACTIONS on Information},
title={On Kernel Parameter Selection in Hilbert-Schmidt Independence Criterion},
year={2012},
volume={E95-D},
number={10},
pages={2564-2567},
abstract={The Hilbert-Schmidt independence criterion (HSIC) is a kernel-based statistical independence measure that can be computed very efficiently. However, it requires us to determine the kernel parameters heuristically because no objective model selection method is available. Least-squares mutual information (LSMI) is another statistical independence measure that is based on direct density-ratio estimation. Although LSMI is computationally more expensive than HSIC, LSMI is equipped with cross-validation, and thus the kernel parameter can be determined objectively. In this paper, we show that HSIC can actually be regarded as an approximation to LSMI, which allows us to utilize cross-validation of LSMI for determining kernel parameters in HSIC. Consequently, both computational efficiency and cross-validation can be achieved.},
keywords={},
doi={10.1587/transinf.E95.D.2564},
ISSN={1745-1361},
month={October},}
Copy
TY - JOUR
TI - On Kernel Parameter Selection in Hilbert-Schmidt Independence Criterion
T2 - IEICE TRANSACTIONS on Information
SP - 2564
EP - 2567
AU - Masashi SUGIYAMA
AU - Makoto YAMADA
PY - 2012
DO - 10.1587/transinf.E95.D.2564
JO - IEICE TRANSACTIONS on Information
SN - 1745-1361
VL - E95-D
IS - 10
JA - IEICE TRANSACTIONS on Information
Y1 - October 2012
AB - The Hilbert-Schmidt independence criterion (HSIC) is a kernel-based statistical independence measure that can be computed very efficiently. However, it requires us to determine the kernel parameters heuristically because no objective model selection method is available. Least-squares mutual information (LSMI) is another statistical independence measure that is based on direct density-ratio estimation. Although LSMI is computationally more expensive than HSIC, LSMI is equipped with cross-validation, and thus the kernel parameter can be determined objectively. In this paper, we show that HSIC can actually be regarded as an approximation to LSMI, which allows us to utilize cross-validation of LSMI for determining kernel parameters in HSIC. Consequently, both computational efficiency and cross-validation can be achieved.
ER -