It has been reported repeatedly that discriminative learning of distance metric boosts the pattern recognition performance. Although the ITML (Information Theoretic Metric Learning)-based methods enjoy an advantage that the Bregman projection framework can be applied for optimization of distance metric, a weak point of ITML-based methods is that the distance threshold for similarity/dissimilarity constraints must be determined manually, onto which the generalization performance is sensitive. In this paper, we present a new formulation of metric learning algorithm in which the distance threshold is optimized together. Since the optimization is still in the Bregman projection framework, the Dykstra algorithm can be applied for optimization. A nonlinear equation has to be solved to project the solution onto a half-space in each iteration. We have developed an efficient technique for projection onto a half-space. We empirically show that although the distance threshold is automatically tuned for the proposed metric learning algorithm, the accuracy of pattern recognition for the proposed algorithm is comparable, if not better, to the existing metric learning methods.
Rachelle RIVERO
Gunma University, Graduate School of Science and Technology,University of the Philippines
Yuya ONUMA
Gunma University, Graduate School of Science and Technology
Tsuyoshi KATO
Gunma University, Graduate School of Science and Technology,Gunma University,Waseda University
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
Rachelle RIVERO, Yuya ONUMA, Tsuyoshi KATO, "Threshold Auto-Tuning Metric Learning" in IEICE TRANSACTIONS on Information,
vol. E102-D, no. 6, pp. 1163-1170, June 2019, doi: 10.1587/transinf.2018EDP7145.
Abstract: It has been reported repeatedly that discriminative learning of distance metric boosts the pattern recognition performance. Although the ITML (Information Theoretic Metric Learning)-based methods enjoy an advantage that the Bregman projection framework can be applied for optimization of distance metric, a weak point of ITML-based methods is that the distance threshold for similarity/dissimilarity constraints must be determined manually, onto which the generalization performance is sensitive. In this paper, we present a new formulation of metric learning algorithm in which the distance threshold is optimized together. Since the optimization is still in the Bregman projection framework, the Dykstra algorithm can be applied for optimization. A nonlinear equation has to be solved to project the solution onto a half-space in each iteration. We have developed an efficient technique for projection onto a half-space. We empirically show that although the distance threshold is automatically tuned for the proposed metric learning algorithm, the accuracy of pattern recognition for the proposed algorithm is comparable, if not better, to the existing metric learning methods.
URL: https://global.ieice.org/en_transactions/information/10.1587/transinf.2018EDP7145/_p
Copy
@ARTICLE{e102-d_6_1163,
author={Rachelle RIVERO, Yuya ONUMA, Tsuyoshi KATO, },
journal={IEICE TRANSACTIONS on Information},
title={Threshold Auto-Tuning Metric Learning},
year={2019},
volume={E102-D},
number={6},
pages={1163-1170},
abstract={It has been reported repeatedly that discriminative learning of distance metric boosts the pattern recognition performance. Although the ITML (Information Theoretic Metric Learning)-based methods enjoy an advantage that the Bregman projection framework can be applied for optimization of distance metric, a weak point of ITML-based methods is that the distance threshold for similarity/dissimilarity constraints must be determined manually, onto which the generalization performance is sensitive. In this paper, we present a new formulation of metric learning algorithm in which the distance threshold is optimized together. Since the optimization is still in the Bregman projection framework, the Dykstra algorithm can be applied for optimization. A nonlinear equation has to be solved to project the solution onto a half-space in each iteration. We have developed an efficient technique for projection onto a half-space. We empirically show that although the distance threshold is automatically tuned for the proposed metric learning algorithm, the accuracy of pattern recognition for the proposed algorithm is comparable, if not better, to the existing metric learning methods.},
keywords={},
doi={10.1587/transinf.2018EDP7145},
ISSN={1745-1361},
month={June},}
Copy
TY - JOUR
TI - Threshold Auto-Tuning Metric Learning
T2 - IEICE TRANSACTIONS on Information
SP - 1163
EP - 1170
AU - Rachelle RIVERO
AU - Yuya ONUMA
AU - Tsuyoshi KATO
PY - 2019
DO - 10.1587/transinf.2018EDP7145
JO - IEICE TRANSACTIONS on Information
SN - 1745-1361
VL - E102-D
IS - 6
JA - IEICE TRANSACTIONS on Information
Y1 - June 2019
AB - It has been reported repeatedly that discriminative learning of distance metric boosts the pattern recognition performance. Although the ITML (Information Theoretic Metric Learning)-based methods enjoy an advantage that the Bregman projection framework can be applied for optimization of distance metric, a weak point of ITML-based methods is that the distance threshold for similarity/dissimilarity constraints must be determined manually, onto which the generalization performance is sensitive. In this paper, we present a new formulation of metric learning algorithm in which the distance threshold is optimized together. Since the optimization is still in the Bregman projection framework, the Dykstra algorithm can be applied for optimization. A nonlinear equation has to be solved to project the solution onto a half-space in each iteration. We have developed an efficient technique for projection onto a half-space. We empirically show that although the distance threshold is automatically tuned for the proposed metric learning algorithm, the accuracy of pattern recognition for the proposed algorithm is comparable, if not better, to the existing metric learning methods.
ER -