The original paper is in English. Non-English content has been machine-translated and may contain typographical errors or mistranslations. ex. Some numerals are expressed as "XNUMX".
Copyrights notice
The original paper is in English. Non-English content has been machine-translated and may contain typographical errors or mistranslations. Copyrights notice
L’apprentissage métrique vise à générer des vecteurs de caractéristiques de faible dimension préservés par similarité à partir d’images d’entrée. La plupart des méthodes d’apprentissage métrique approfondi supervisé existantes définissent généralement une fonction de perte soigneusement conçue pour imposer une contrainte sur la position relative entre les échantillons dans un espace dimensionnel inférieur projeté. Dans cet article, nous proposons une nouvelle architecture appelée Naive Similarity Discriminator (NSD) pour apprendre la distribution d'échantillons faciles et prédire leur probabilité d'être similaires. Notre objectif consiste à encourager le réseau de générateurs à générer des vecteurs dans des positions appropriées dont la similarité peut être distinguée par notre discriminateur. Des expériences de comparaison adéquates ont été réalisées pour démontrer la capacité de notre modèle proposé sur les tâches de récupération et de regroupement, avec une précision dans un rayon spécifique, des informations mutuelles normalisées et F1 score comme mesure d’évaluation.
Yi-ze LE
Chongqing University
Yong FENG
Chongqing University
Da-jiang LIU
Chongqing University
Bao-hua QIANG
Guilin University of Electronic Technology
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.
Copier
Yi-ze LE, Yong FENG, Da-jiang LIU, Bao-hua QIANG, "Adversarial Metric Learning with Naive Similarity Discriminator" in IEICE TRANSACTIONS on Information,
vol. E103-D, no. 6, pp. 1406-1413, June 2020, doi: 10.1587/transinf.2019EDP7278.
Abstract: Metric learning aims to generate similarity-preserved low dimensional feature vectors from input images. Most existing supervised deep metric learning methods usually define a carefully-designed loss function to make a constraint on relative position between samples in projected lower dimensional space. In this paper, we propose a novel architecture called Naive Similarity Discriminator (NSD) to learn the distribution of easy samples and predict their probability of being similar. Our purpose lies on encouraging generator network to generate vectors in fitting positions whose similarity can be distinguished by our discriminator. Adequate comparison experiments was performed to demonstrate the ability of our proposed model on retrieval and clustering tasks, with precision within specific radius, normalized mutual information and F1 score as evaluation metrics.
URL: https://global.ieice.org/en_transactions/information/10.1587/transinf.2019EDP7278/_p
Copier
@ARTICLE{e103-d_6_1406,
author={Yi-ze LE, Yong FENG, Da-jiang LIU, Bao-hua QIANG, },
journal={IEICE TRANSACTIONS on Information},
title={Adversarial Metric Learning with Naive Similarity Discriminator},
year={2020},
volume={E103-D},
number={6},
pages={1406-1413},
abstract={Metric learning aims to generate similarity-preserved low dimensional feature vectors from input images. Most existing supervised deep metric learning methods usually define a carefully-designed loss function to make a constraint on relative position between samples in projected lower dimensional space. In this paper, we propose a novel architecture called Naive Similarity Discriminator (NSD) to learn the distribution of easy samples and predict their probability of being similar. Our purpose lies on encouraging generator network to generate vectors in fitting positions whose similarity can be distinguished by our discriminator. Adequate comparison experiments was performed to demonstrate the ability of our proposed model on retrieval and clustering tasks, with precision within specific radius, normalized mutual information and F1 score as evaluation metrics.},
keywords={},
doi={10.1587/transinf.2019EDP7278},
ISSN={1745-1361},
month={June},}
Copier
TY - JOUR
TI - Adversarial Metric Learning with Naive Similarity Discriminator
T2 - IEICE TRANSACTIONS on Information
SP - 1406
EP - 1413
AU - Yi-ze LE
AU - Yong FENG
AU - Da-jiang LIU
AU - Bao-hua QIANG
PY - 2020
DO - 10.1587/transinf.2019EDP7278
JO - IEICE TRANSACTIONS on Information
SN - 1745-1361
VL - E103-D
IS - 6
JA - IEICE TRANSACTIONS on Information
Y1 - June 2020
AB - Metric learning aims to generate similarity-preserved low dimensional feature vectors from input images. Most existing supervised deep metric learning methods usually define a carefully-designed loss function to make a constraint on relative position between samples in projected lower dimensional space. In this paper, we propose a novel architecture called Naive Similarity Discriminator (NSD) to learn the distribution of easy samples and predict their probability of being similar. Our purpose lies on encouraging generator network to generate vectors in fitting positions whose similarity can be distinguished by our discriminator. Adequate comparison experiments was performed to demonstrate the ability of our proposed model on retrieval and clustering tasks, with precision within specific radius, normalized mutual information and F1 score as evaluation metrics.
ER -